JP5164880B2 - Image processing apparatus, image processing method, and image display apparatus - Google Patents

Description

  The present invention relates to an image processing apparatus and an image processing method for generating an output image signal composed of data in which the number of gradations is expanded by extending the number of gradations of data of each pixel of an input image signal, and the image The present invention relates to an image display device to which a processing device or an image processing method is applied.

  Devices such as liquid crystal, plasma, EL (electroluminescence), DMD (digital mirror device), etc., display images by modulating matrix pixels arranged discretely by reflection or transmission of light or light emission. It is used in various image display devices (displays) such as flat-screen televisions, projection televisions, projectors, and computer monitors. In recent years, high-definition broadcasting and computer processing speed have been improved, and the display has been improved in definition, and display has been rapidly increased in brightness due to device improvements. On the other hand, the number of gradations (bits) that can be displayed is expanded as the display brightness increases, so users can identify subtle gradation differences between pixels when displaying digitized image signals. Quantization noise has become prominent.

  When receiving an analog image signal, the number of bits of the digital signal after conversion by the analog-to-digital converter is increased (for example, the number of bits that was conventionally 8 bits is increased to 10 bits), and quantization is performed. It is possible to reduce noise. However, in a display capable of displaying a high-definition image, increasing the speed of the analog-digital converter and increasing the number of quantizations involve a significant cost increase.

  When a digital image signal is received (for example, in the case of digital television broadcasting), an image signal having a quantization number determined by an agreement between a transmission side and a reception side in digital transmission is transmitted / received (generally, 8 By performing bit quantization, an 8-bit gradation image signal is generated), and the quantization number cannot be changed unless the agreement is changed. Similarly, since the number of quantizations is also determined in advance for digital contents on DVDs and networks, the number of quantizations of the contents themselves cannot be increased unless a new standard is provided.

  Furthermore, in image signal processing such as television, in order to display high-impact video, processing to expand the dynamic range of the received image signal and conversion to high-contrast video by changing the tone-brightness conversion characteristics Processing to be performed. When such gradation correction processing is performed, a high-contrast video can be displayed, while gradation skip (gradation jump) occurs due to gradation correction, and quantization noise increases.

  In such a case, as a method of reducing the quantization noise of the image signal by digital image processing, a gradation conversion technique (multi-gradation technique) that increases the number of gradations of the image signal by adding lower bits to the image signal. ) As this gradation conversion technique, a technique has been proposed in which a signal having a specific gradation change in an image signal is detected and the number of gradations is expanded for the detected necessary pixels.

  For example, Japanese Patent Laid-Open No. 2003-304400 (Patent Document 1) uses an image memory to detect the same luminance region that is a set of adjacent pixels at the same luminance level in an image signal from the image memory, A boundary between adjacent identical luminance regions based on the determination result of the false contour detector that determines whether the difference in luminance level between adjacent identical luminance regions matches a preset value or not An image processing apparatus including a pixel value converter that converts a luminance level of a pixel in the vicinity to a halftone is disclosed.

  Further, in Japanese Patent No. 3710131 (Patent Document 2), the detection means has a low-frequency portion of the input image signal from the line memory (a plurality of first pixel values are continuous, and a predetermined level pixel from the first pixel value). The second pixel value having a changed value is detected in a plurality of consecutive portions), and the signal extension means adds a predetermined number of bits to the bit string of the first or / and second pixel value in the predetermined range of the low frequency portion. An image processing apparatus is disclosed that performs the extended correction so that the pixel values sequentially change.

  Further, Japanese Patent Laid-Open No. 2007-158847 (Patent Document 3) sequentially inputs video signals to a shift register, and uses a pixel signal held in the register so that the signal level difference is within a preset signal level difference. When the signal level difference is, for example, 1 bit, a continuous pixel equal to the signal level of the pixel immediately before the signal level difference is determined is determined. On the other hand, there has been proposed a gradation conversion device that sets the amount of change in signal level between pixels by averaging, and generates a video signal by assigning bits based on the set amount of change.

  Furthermore, Japanese Patent Application Laid-Open No. 2007-181189 (Patent Document 4) discloses an apparatus and method for making it less susceptible to noise components included in an image when detecting a change in gradation in a moderate gradation image area. Has proposed. In Patent Literature 4, when detecting a linear interpolation application region by sequentially scanning a pixel row of an input image signal, a change in gradation within a predetermined range (for example, a minimum gradation difference) is detected. If the tone difference at the position next to the position where the tone change is detected is the same as the tone difference at the position before the position where the tone change is detected, the tone change is due to noise or the like. It is disclosed that the gradation difference of the pixel in which the gradation change is detected is determined as the gradation value of the pixel at the position before and after the pixel.

JP 2003-304400 A (summary, FIG. 1) Japanese Patent No. 3710131 (FIGS. 2-4) JP 2007-158847 A (summary, FIGS. 2 and 12) JP 2007-181189 A (Summary, FIGS. 3, 6, 7 and 9)

  As described above, in an image signal in a television broadcast, a DVD, digital content on a network, or the like, a signal having a predetermined quantization number (bit number) is input to the image processing apparatus. When the number of bits that can be displayed on the display is smaller or when a tone jump occurs due to image processing such as tone correction processing, the tone changes very slowly because there is no image outline (edge). A stripe pattern that does not originally exist in the image area is perceived as a pseudo contour (also referred to as “pseudo contour”). For this reason, in the devices of Patent Documents 1 to 4, detection of a region where the same gradation level continues and detection of a change in gradation level where the gradation difference is within a certain value are performed in an image memory (for example, Can store the gradation level for the pixel width of the continuous area you want to detect), suppress the pseudo contour for the gradation change of the area specified based on the detection result, and smooth the gradation change Thus, the number of gradations is expanded by linearly interpolating between the gradation change positions to increase the number of bits of the image signal. As described above, since the level of gradation change is conventionally detected using the image memory, there is a problem that the circuit scale of the image processing apparatus increases and the apparatus becomes expensive. In addition, it is required to perform processing in real time for image signals in which images are sent continuously in time, such as a television, but the above image processing performed using an image memory is performed in real time. There was a problem that it was difficult to do.

  In addition, in the devices disclosed in Patent Documents 1 to 4, noise such as noise and error, such as a lamp (Ramp) image created by computer graphics or an image input as a digital image signal, for example. Appropriate detection of areas where pixels with the same gradation value continue, and changes in gradation where the gradation difference is within a certain value, for images with few components and gradation changes monotonously (increase or decrease) It can be detected. However, in the apparatuses disclosed in Patent Documents 1 to 4, in the case of an image signal including a minute gradation change due to a noise component, such as a change due to image detail or dither noise, or an analog image signal to a digital image. In the case of converted image signals, etc., the gradation change due to the noise component adversely affects the detection of flat areas where pixels with the same gradation value are continuous, or the detection of small gradation changes where the gradation difference is within a certain value. There is a problem that false detection occurs. In addition, in the devices disclosed in Patent Documents 1 to 4, the gradation number expansion process is performed by linearly interpolating the gradation value between the pixels at the gradation change position with respect to the gradation change in the specified region. . For this reason, although noise components are suppressed, interpolation processing is also performed on the pixels that are noise components, and conversion is performed so that the change in gradation is smooth. Sharpness (contour information) is not maintained, image blurring occurs, quantization noise or pseudo contour cannot be reduced well, and image quality deteriorates.

  Furthermore, in the apparatus described in Patent Document 4, when a gradation change within a predetermined range is detected, the gradation difference at the position next to the position where the gradation change is detected detects the gradation change. If it is the same as the gradation difference at the position before the position, it is determined that the gradation change is due to the noise component, and the gradation difference of the pixel where the gradation change is detected is determined as the position before and after the pixel. As a configuration in which processing that is regarded as a gradation value of a pixel is added, interpolation processing that suppresses noise components is performed, and gradation number expansion processing is performed. However, the gradation change due to the noise component does not necessarily have the same gradation value before and after the gradation change position, and the magnitude of the gradation change (gradation difference) may be a certain value or more. Therefore, there is a problem that noise components may not be detected properly.

  Therefore, the present invention has been made to solve the above-described problems of the prior art, and the purpose thereof is to greatly increase the circuit scale even when a noise component is included in the input image signal. An output image that displays an image in a section where gradation values change very slowly by real-time signal processing as a high-quality image with smooth gradation changes while suppressing image quality degradation due to noise components An object is to provide an image processing apparatus and an image processing method capable of generating a signal.

  Another object of the present invention is that even if the input image signal includes a noise component, the gradation value changes very slowly by real-time signal processing without significantly increasing the circuit scale. It is an object of the present invention to provide an image display apparatus capable of displaying an image in a section to be displayed as a high-quality image with smooth gradation change while suppressing deterioration in image quality due to noise components.

  The image processing apparatus according to the present invention receives an input image signal composed of discrete data in the time direction and the space direction, and generates the input image signal by conversion processing that expands the number of gradations and corrects the gradation value. An image processing apparatus having a function of outputting an output image signal composed of the converted data, wherein a gradation change amount that is a difference between gradation values of pixels arranged adjacent to each other is calculated from the input image signal. Based on the gradation change amount detection means to be obtained and the gradation change amount, each pixel is divided into a pixel belonging to a gradation flat region in which the value of the gradation change amount is 0, and a noise component included in the input image signal To determine whether the pixel belongs to a noise area where the value of the gradation change amount is not 0, or a pixel in a non-noise gradation change position that is an area other than the gradation flat area and the noise area. , The judgment A pixel area determination unit that generates an area signal indicating the result, and a gradation between the reference pixel that is the pixel at the non-noise gradation change position and the pixel at the next non-noise gradation change position that receives the area signal. A flat / noise area width data predicted value indicating a predicted value of a flat / noise area width that is a width of a flat / noise area composed of a flat area and a noise area, the flat / noise immediately before the reference pixel The flatness predicting the width of the flat / noise region immediately after the reference pixel based on the region width and the previous flat / noise region width data prediction value generated at the immediately preceding reference pixel detected immediately before the reference pixel. The flatness immediately after the reference pixel is calculated based on the flatness / noise area width determination means for generating the predicted noise area width data value, and the gradation change amount and the predicted flatness / noise area width data value. The area signal indicates the inclination detection means for obtaining the inclination data value used when setting the inclination of the gradation value of the converted data in the noise area, and the gradation change amount and the inclination data value. A correction bit generation unit that obtains a correction value of the gradation value of the input image signal according to the determination result and generates correction bit data corresponding to the correction value, and according to the correction bit data for the gradation value Pixel value calculation means for generating the output image signal composed of the converted data by subtracting or adding the obtained values.

  In another image processing apparatus of the present invention, an input image signal composed of discrete data in the time direction and the spatial direction is input, the number of gradations of the input image signal is expanded, and the gradation value is corrected. An image processing apparatus having a function of outputting an output image signal composed of post-conversion data generated by a conversion process, wherein the level is a difference between gradation values of pixels arranged adjacent to each other from the input image signal. Tone change amount detecting means for obtaining a tone change amount, and based on the tone change amount, each pixel is converted into a pixel belonging to a tone flat region where the tone change amount is 0, and the input image signal. The pixel is classified into one of a pixel belonging to a noise area where the value of the gradation change amount is not 0 by a noise component included, and a pixel at a non-noise gradation change position that is an area other than the flat gradation area and the noise area. Make a decision to A pixel region determination unit that generates a region signal indicating the result of the determination; and a region between the reference pixel that is the pixel at the non-noise gradation change position and the pixel at the next non-noise gradation change position that receives the region signal. The flat / noise area width is measured by measuring the flat / noise area width indicating the flat / noise area width, which is the width of the flat / noise area consisting of the gray level flat area and the noise area in FIG. Used when setting the gradient of the gradation value of the converted data in the flat / noise area immediately after the reference pixel based on the determination means, the gradation change amount, and the measured flat / noise area width. Using the inclination detection means for obtaining the inclination data value, the gradation change amount and the inclination data value, a correction value of the gradation value of the input image signal corresponding to the determination result indicated by the area signal is obtained, To the correction value Correction bit generation means for generating corresponding correction bit data, and a value corresponding to the correction bit data is subtracted or added to the gradation value to generate the output image signal composed of the converted data. And a pixel value calculation means.

  Furthermore, the image display apparatus of the present invention includes display means for displaying an image signal, and causes the display means to display an image signal based on the image signal whose number of gradations is expanded by the image processing apparatus.

  According to the image processing apparatus and the image processing method of the present invention, even when a noise component is included in the input image signal, the level of the image signal is changed in the interval in which the gradation value changes very slowly with respect to the input image signal. Since the output image signal is generated by performing gradation conversion processing that expands the logarithm and corrects the gradation value appropriately, the circuit scale is not increased significantly, and real-time signal processing makes it very gentle. Output image signal that can display an image in a section where the gradation value changes as a high-quality image with smooth gradation changes without losing image details or loss of sharpness. There is an effect that can be generated.

  Further, according to the image display device of the present invention, even when a noise component is included in the input image signal, the number of gradations of the image signal in the interval in which the gradation value changes very gently with respect to the input image signal. Since the output image signal is generated by performing the gradation conversion processing that appropriately corrects the gradation value, and the image based on the output image signal is displayed, the interval in which the gradation value changes very slowly Thus, the image can be displayed as a high-quality image with a smooth gradation change without losing image detail or loss of sharpness.

It is a block diagram which shows roughly an example of a structure of the image processing apparatus by the Embodiment 1 of this invention (namely, apparatus which can implement the image processing method by Embodiment 1). (A)-(c) is a figure which shows an example of the signal waveform (when the gradation value of an input image signal raises gradually, and a noise component is not included) in the image processing apparatus by Embodiment 1, ( a) shows the gradation value near the least significant bit of the 8-bit input image signal, (b) shows the gradation change amount and gradation change position of the input image signal, and (c) shows the gradation value. The gradation value near the least significant bit of the 12-bit output image signal whose number is expanded and the gradation value is corrected is shown. (A)-(c) is a figure which shows an example of the signal waveform (When the gradation value of an input image signal raises gradually, and a noise component is included) in the image processing apparatus by Embodiment 1, (a ) Indicates the gradation value near the least significant bit of the 8-bit input image signal, (b) indicates the gradation change amount and gradation change position of the input image signal, and (c) indicates the number of gradations. Indicates a gradation value in the vicinity of the least significant bit of the 12-bit output image signal with the gradation value corrected and the gradation value corrected. (A) And (b) is a figure which shows an example of the pixel determined to be a noise component in Embodiment 1, (a) has a gradation value increased 1 gradation from the pixel immediately before and immediately after. The noise component consisting of one pixel (one “one gradation changing pixel”) is shown, and (b) is one element whose gradation value is reduced by one gradation from the immediately preceding and immediately following pixels. The noise component consisting of this pixel (one 'one gradation changing pixel') is shown. It is a block diagram which shows roughly an example of a structure of the pixel area determination means shown by FIG. (A)-(g) is a figure for demonstrating an example of operation | movement of the pixel area determination means shown by FIG.1 and FIG.5. It is a block diagram which shows roughly an example of a structure of the flat and noise area | region width determination means shown by FIG. It is a block diagram which shows roughly an example of a structure of the edge change detection means shown by FIG. It is a figure which shows an example of operation | movement of the edge change detection means shown by FIG.1 and FIG.8. (A) And (b) is a figure which shows the example of the gradation change detected by the edge change detection means shown by FIG.1 and FIG.8. It is a block diagram which shows roughly an example of a structure of the inclination detection means shown by FIG. (A) And (b) is a conceptual diagram for demonstrating the inclination data value calculated by the inclination detection means shown by FIG.1 and FIG.9. It is a block diagram which shows roughly an example of a structure of the correction | amendment bit production | generation means shown by FIG. It is a figure for demonstrating operation | movement of the pixel value calculating means shown by FIG. 5 is a flowchart illustrating an operation (part 1) of the image processing apparatus according to the first embodiment. 6 is a flowchart illustrating an operation (part 2) of the image processing apparatus according to the first embodiment. It is a flowchart which shows operation | movement of the pixel area determination means shown by FIG. (A)-(k) is a figure for demonstrating an example (when a gradation value rises very gently and a noise component is included) of the operation | movement of the image processing apparatus by Embodiment 1. FIG. (A)-(k) is a figure for demonstrating the other example (when a gradation value falls very gently and a noise component is included) of the operation | movement of the image processing apparatus by Embodiment 1. FIG. It is a figure for demonstrating the other example of operation | movement of the pixel value calculating means shown by FIG. It is a block diagram which shows roughly an example of a structure of the image processing apparatus by the Embodiment 2 of this invention (namely, apparatus which can implement the image processing method by Embodiment 2). It is a block diagram which shows roughly an example of a structure of the pixel area determination means shown by FIG. (A) And (b) is a figure which shows an example of the pixel determined to be a noise component in Embodiment 2, (a) has a gradation value increased 1 gradation from the pixel immediately before and immediately after. Shows a noise component composed of two consecutive pixels (two consecutive “1 gradation changing pixels”), and (b) shows that the gradation value is reduced by one gradation compared to the immediately preceding and immediately following pixels. A noise component composed of two consecutive pixels (two consecutive “1 gradation changing pixels”) is shown. It is a block diagram which shows roughly an example of a structure of the image processing apparatus (namely, apparatus which can implement the image processing method by Embodiment 3) by Embodiment 3 of this invention. It is a block diagram which shows roughly an example of a structure of the pixel area determination means shown by FIG. (A)-(e) is a figure for demonstrating an example of operation | movement of the pixel area determination means shown by FIG. It is a flowchart which shows operation | movement of the pixel area determination means shown by FIG. FIG. 25 is a block diagram schematically illustrating another example of the configuration of the pixel region determination unit illustrated in FIG. 24. It is a block diagram which shows roughly an example of a structure of the image processing apparatus (namely, apparatus which can implement the image processing method by Embodiment 4) by Embodiment 4 of this invention. (A) And (b) is a conceptual diagram for demonstrating the inclination data value calculated by the inclination detection means shown by FIG. It is a block diagram which shows roughly an example of a structure of the correction | amendment bit production | generation means shown by FIG. FIG. 30 is a block diagram schematically illustrating another example of the configuration of the correction bit generation unit illustrated in FIG. 29. 14 is a flowchart illustrating an operation (No. 1) of the image processing apparatus according to the fourth embodiment. 14 is a flowchart illustrating an operation (part 2) of the image processing apparatus according to the fourth embodiment. (A)-(k) is a figure for demonstrating an example (when a gradation value rises very gently and a noise component is included) of the operation | movement of the image processing apparatus by Embodiment 4. FIG. (A)-(k) is a figure for demonstrating the other example (when a gradation value falls very gently and a noise component is included) of the operation | movement of the image processing apparatus by Embodiment 4. FIG. 35 is a flowchart showing another example of the operation of the image processing apparatus according to the fourth embodiment (an example instead of FIG. 34). It is a block diagram which shows an example of a structure of the image processing apparatus (namely, apparatus which can implement the image processing method by Embodiment 5) by Embodiment 5 of this invention. It is a block diagram which shows roughly an example of a structure of the image display apparatus by Embodiment 6 of this invention. It is a block diagram which shows roughly an example of a structure of the image display apparatus by Embodiment 7 of this invention.

<< 1 >> Embodiment 1
<< 1-1 >> Overview.
FIG. 1 is a block diagram schematically showing a configuration of an image processing apparatus according to Embodiment 1 of the present invention (that is, an apparatus capable of performing the image processing method according to Embodiment 1). The image processing apparatus according to the first embodiment expands the number of bits (the number of gradations) for converting an m-bit (m is a natural number) digital input image signal Da into an n-bit (n is a natural number) digital output image signal Db. A gradation conversion unit 1 capable of performing gradation conversion processing (n> m) is provided. The gradation conversion processing by the gradation converting means 1 is performed by, for example, expanding the number of bits and adding the correction bit data Cd, which is bit data lower than the input image signal Da, to the m-bit input image signal Da. This is a process of correcting the tone value and generating an n-bit output image signal Db.

  The gradation converting means 1 shown in FIG. 1 obtains the difference in gradation value between pixels in the spatial direction of the input image signal Da as the gradation change amount df when the gradation changes, and the obtained gradation change. Based on the amount df, each pixel is divided into a pixel belonging to a “gradation flat region” (Ar = 0) where the gradation change amount df is 0, and a gradation change amount df depending on a noise component included in the input image signal Da. Pixels belonging to a “noise area” (also referred to as “noise gradation change area”) where the value becomes a value other than 0, and a “non-noise gradation change position” that is neither a flat gradation area nor a noise area ( It is determined that the pixel is classified into any one of “non-noise gradation changing region”), and an area signal Ar indicating the result of the determination is output. That is, the gradation converting means 1 performs a determination for dividing the pixel area into three areas, a flat gradation area, a noise area, and a non-noise gradation change position, and an area signal Ar indicating the determination result is obtained. Output. In the first embodiment, the region signal Ar for the pixel in the flat gradation region (df = 0) has the value “0” (Ar = 0), and the region signal Ar for the pixel in the noise region has the value “1”. (Ar = 1), and the region signal Ar for the pixel at the non-noise gradation change position has a value “2” (Ar = 2). However, as the value of the area signal Ar, other values may be adopted as long as each pixel indicates a pixel belonging to which of the three areas. Further, the present invention includes a case where the gray level flat area and the noise area are treated as one common area (a “flat / noise area” described later), and therefore the common area and the non-noise level are used as the area signal Ar. It is also possible to adopt a value indicating two areas of the tone change position.

  Further, the gradation converting means 1 is a pixel at the non-noise gradation change position (Ar = 2) that appears next from the reference pixel that is a pixel at the non-noise gradation change position (Ar = 2) based on the area signal Ar. The gradation flat area and noise area (the gradation flat area and noise area (Ar = 0, 1) are referred to as “flat / noise area”), A “flat / noise area width data predicted value” indicating the predicted value is generated. The flat / noise area width data prediction value dst obtained by predicting the width of the flat / noise area immediately after the reference pixel is generated in the flat / noise area width immediately before the reference pixel and the immediately preceding reference pixel detected immediately before the reference pixel. Is calculated based on the predicted value immediately before the flat / noise area width data (for example, calculation of “(0 + 7) / 2”, calculation of “(4 + 7) / 2” in FIG. 18F described later), and the like). .

  Further, the tone converting means 1 determines the non-noise tone change position based on the amplitude limit tone change amount Xlm obtained by limiting the amplitude of the tone change amount df and the flat / noise area width data predicted value dst. An inclination data value Ksl used when setting the inclination of the gradation value of the converted data in the flat / noise area immediately after the reference pixel of (Ar = 2) is obtained, and the edge of the image (based on the gradation change amount df) is obtained. A correction limiting coefficient rc for limiting the gradient of the gradation value of the output image signal due to the contour) is set, and by using the amplitude limited gradation change amount Xlm, the gradient data value Ksl, and the correction limiting coefficient rc, the input image signal Da A correction value of the gradation value La (x) is obtained, correction bit data Cd corresponding to the correction value is generated, and correction bit data Cd is generated for pixel data of the gradation value La (x) of the input image signal Da. Subtract or add the corresponding value Te, and it generates an output image signal Db composed of converted data.

  According to the image processing apparatus or the image processing method of the first embodiment, even if the input image signal Da includes a noise component, the noise region (Ar = 1) is detected as the non-noise gradation change position (Ar = 2), and a flat / noise area width in a section where the gradation value of the input image signal Da changes very slowly (also referred to as “gradation flat area width” when there is no noise component). Can be accurately detected without being affected by the noise component. Therefore, the slope data value Ksl calculated using the flat / noise region width data predicted value dst can be appropriately obtained without being affected by the noise component, and the correction bit calculated using the slope data value Ksl. The data Cd can be obtained appropriately, and the gradation conversion processing of the image signal performed using the correction bit data Cd can be appropriately performed. As a result, an image in a section in which the gradation value changes very slowly in the input image signal Da is high quality with smooth gradation change without losing image details and loss of sharpness. An output image signal Db that can be displayed as a simple image can be generated. In addition, since the processing of the gradation converting means 1 in the first embodiment does not require the use of an image memory having a capacity corresponding to the flat / noise area width, the gradation converting process does not significantly increase the circuit scale. Can be realized by real-time signal processing.

<< 1-2 >> Configuration.
<< 1-2-1 >> Image processing apparatus.
As shown in FIG. 1, the image processing apparatus according to the first embodiment receives an input image signal Da composed of discrete data in the time direction and the spatial direction, and expands the number of gradations of the input image signal Da. In addition, a gradation conversion unit 1 is provided that outputs an output image signal Db composed of post-conversion data generated by a conversion process for correcting gradation values (gradation levels). The gradation converting means 1 includes a gradation change amount detecting means 10, a pixel area determining means 20, a flat / noise area width determining means 30, an inclination detecting means 40, an edge change detecting means 60, and a correction bit generating means. 50, delay compensation means 91, and pixel value calculation means 92.

  The gradation change amount detection means 10 calculates, from the input image signal Da, a gradation change amount df that is a difference between gradation values of pixels adjacently arranged (for example, adjacently arranged in the horizontal direction or the vertical direction). A gradation change amount calculating means 11 for outputting, and a gradation change amount limiting means 12 for generating and outputting an amplitude limited gradation change amount Xlm in which the value of the gradation change amount df is restricted within a predetermined range are provided. Yes.

  The pixel area determination unit 20 sets each pixel from the gradation change amount df from the gradation change amount detection unit 10 to “pixel belonging to a flat gradation region” where df = 0, and df ≠ 0 due to a noise component. The determination is made to classify the pixel into either “a pixel belonging to the noise region” or “a pixel at the non-noise gradation change position” where df ≠ 0 regardless of the noise component, and an area signal Ar indicating the determination result is output. By performing the determination to classify (discriminate) each pixel into three regions in this way, a pixel region constituted by pixels arranged in the horizontal direction (or vertical direction) is a “gradation flat region” (Ar = 0), It is distinguished (divided) into three areas of “noise area” (Ar = 1) and “non-noise gradation change position” (Ar = 2).

  The flat / noise area width determining unit 30 detects the detected non-noise gradation every time a pixel at the non-noise gradation changing position (Ar = 2) is detected from the area signal Ar from the pixel area determining unit 20. The pixel at the change position is set as a reference pixel, and the flatness / noise area width of the flat / noise area immediately after this reference pixel (that is, the reference pixel at the non-noise gradation change position and the pixel at the non-noise gradation change position appearing next) Is a flat / noise region width data predicted value dst). The flat / noise area width determination means 30 calculates the flatness / noise area width of the flat / noise area immediately before the reference pixel and the previous flat / noise area width data predicted value generated in the immediately preceding reference pixel detected immediately before the reference pixel. Based on the above, a flat / noise area width data prediction value dst in which the flat / noise area width immediately after the reference pixel is predicted is generated.

  The inclination detection means 40 is based on the amplitude-limited gradation change amount Xlm from the gradation change amount detection means 10 and the flat / noise area width data predicted value dst from the flat / noise area width determination means 30. This is means for obtaining an inclination data value Ksl used when setting the inclination of the gradation value of the converted data in the flat / noise area immediately after the reference pixel at the tone change position.

  The edge change detection means 60 determines whether or not the gradation change at the gradation change position is a gradation change caused by an edge of the image based on the gradation change amount df from the gradation change amount detection means 10. This is means for setting a correction limiting coefficient rc for limiting the gradient of the gradation value of the output image signal Db according to the determination result.

  The correction bit generation means 50 is based on the amplitude limit gradation change amount Xlm from the gradation change amount detection means 10, the inclination data value Ksl from the inclination detection means 40, and the correction restriction coefficient rc from the edge change detection means 60. A means for obtaining a correction value of the gradation value of the input image signal Da and generating correction bit data Cd corresponding to the correction value.

  The delay compensation unit 91 is a unit that delays the input image signal Da and outputs it to the pixel value calculation unit 92 as an image signal Dd. The pixel value calculation unit 92 subtracts or adds a value corresponding to the correction bit data Cd to the gradation value of the input image signal Da (Dd) to generate an output image signal Db composed of the converted data. It is.

  The gradation converting means 1 adds the correction bit data Cd as the lower bits to the m-bit input image signal Da (Dd), thereby converting the m-bit input image signal Da (Dd) into the n-bit output image signal Db. The number of bits is expanded and output sequentially. Further, the gradation conversion means 1 uses the correction limiting coefficient rc to determine that the gradation change amount df at the gradation change position is large (the change in the gradation value is a change due to the edge (outline) of the image. The correction bit data Cd is set to 0 (or close to 0), and when the gradation change amount df at the gradation change position is small, the correction bit data Cd is set to a value based on the inclination data value Ksl.

  The image processing apparatus and the image processing method according to the first embodiment greatly increases the circuit scale for an image region in which a gradation value changes very gently in a certain direction (for example, the horizontal direction) of a screen on which an image is displayed. Without increasing, the gradation change (df ≠ 0) in the noise region (Ar = 1) is detected separately from the gradation change (df ≠ 0) in the non-noise gradation change position (Ar = 2), and Detect tonal changes more appropriately. For this reason, the gradation of the image signal is expanded to a sufficient number of gradations based on the detection result, and the gradation changes smoothly without losing image details or loss of sharpness. An output image signal Db that can display an image can be generated.

  In the image processing apparatus and the image processing method according to the first embodiment, when it is determined that the change in the gradation value at the gradation change position is a change due to the edge (contour) of the image, the correction limiting coefficient rc Thus, the output image signal Db faithful to the value of the input image signal Da can be generated by setting the correction bit data Cd to 0 (or approaching 0). For this reason, the display means (not shown in FIG. 1) to which the output image signal Db whose gradation has been converted by the image processing apparatus of the first embodiment is to be displayed faithfully to the value of the input image signal Da. In the contour, the image can be displayed faithfully to the value of the input image signal Da, and the gradation value of the input image signal Da changes very slowly (the gradation gradation region is the noise region. In addition, the display of a high-quality image with reduced pseudo contour can be realized.

  In the present application, the configuration indicated as “... Means” may be realized by an electric circuit (hardware) or realized by software.

  In FIG. 1, an analog or digital image signal is converted into an m-bit digital image signal format and input to the gradation conversion means 1 as an input image signal Da. For example, the gradation converting means 1 generates Q-bit correction bit data Cd, adds the Q-bit correction bit data Cd to the least significant side of the m-bit input image signal Da, and extends the number of gradations. An n-bit output image signal Db is generated. The bit number m of the input image signal Da is, for example, 8 bits, the bit number Q of the gradation correction value is, for example, 4 bits, and the bit number n of the output image signal Db is, for example, 12 bits. . Further, in the first embodiment, a case will be described in which the fixed direction for detecting the change in the gradation value of the input image signal Da is the horizontal direction of the screen (screen display horizontal scanning direction). Further, in the first embodiment, the gradation converting means 1 has a difference of the least significant 1 bit of the input image signal Da having the gradation change amount df of 8 bits, that is, a difference of gradation values between pixels is +1. Alternatively, an interval in which the gradation value changes very slowly as a section where the gradation value changes −1, and gradation conversion processing is performed, and an image in which the gradation changes smoothly in a section where the gradation value changes very slowly. Can be displayed as

  The image signal has a format represented by luminance and chromaticity or a format represented by primary color signals such as red (R), green (G), and blue (B). The image signal processing can be performed for any type of signal. In the following, a case where the gradation converting unit 1 processes any one type of image signal will be described. In the following, image processing is performed in the horizontal direction of the input image signal Da, gradation conversion correction processing for expanding and correcting the number of gradations of the 8-bit input image signal Da, and 12-bit output image signal is performed. A case of outputting as Db will be described.

  FIGS. 2A to 2C are diagrams showing signal waveforms (when the gradation value of the input image signal Da rises very slowly and does not include noise components) in the image processing apparatus according to the first embodiment. 2A shows the gradation value La (x) (range of gradation values 1 to 6) in the vicinity of the least significant bit of the 8-bit input image signal Da, and FIG. The gradation change amount df and the gradation change position (only the non-noise gradation change position) of the image signal Da are shown, and FIG. 2C shows the lowest order of the 12-bit output image signal Db in which the number of gradations is expanded. A gradation value Lb (x) in the vicinity of a bit (range of gradation values 16 to 96) is shown. 2A to 2C indicate the positions x of pixels arranged in the horizontal direction of the screen. Also, the vertical axis in FIG. 2A indicates the gradation value La (x) of the 8-bit input image signal Da at each pixel position, and the vertical axis in FIG. 2B indicates the gradation change amount df. In FIG. 2C, the vertical axis indicates the gradation value Lb (x) of the 12-bit output image signal Db at each pixel position. In FIG. 2A, the gradation value La (x) of the input image signal Da corresponds to the least significant 1 bit of the gradation value every 8 pixels between the horizontal pixel positions x = 8-40. A case where the gradation gradually increases by one gradation is shown. FIG. 2B shows the gradation change amount df of the input image signal Da obtained by the gradation conversion means 1 and the gradation flat region width (“flat / noise region width”) in which the gradation value does not change. ) And the pixel position. FIG. 2C shows a change in the gradation value Lb (x) of the 12-bit output image signal Db output from the gradation converting means 1. When the gradation difference is large, for example, larger than 2 gradations, it can be said that the gradation level of the signal does not change smoothly at the boundary between adjacent pixels. For this reason, when the gradation difference is large (for example, two gradations or more), there is no problem even if an outline can be seen at the boundary of the gradation. On the other hand, when the gradation difference is 1 gradation, although the gradation level of the signal changes smoothly at the boundary between adjacent pixels, quantization noise or pseudo contour may occur at the boundary. . In such a pixel of an image signal in which the gradation value changes gradually (at a low frequency) with one gradation, the gradation value changes more smoothly (for example, FIG. )), Gradation conversion for increasing the number of gradations of the image signal Da is performed.

  FIGS. 3A to 3C are signal waveforms in the image processing apparatus according to the first embodiment (when the gradation value of the input image signal Da increases very slowly and includes noise components such as noise and error). FIG. 3A shows the gradation value La (x) (range of gradation values 1 to 6) near the least significant bit of the 8-bit input image signal Da, and FIG. ) Shows the gradation change amount df and the gradation change position (in this figure, the noise region and the non-noise gradation change position) of the input image signal Da, and FIG. Further, the gradation value Lb (x) (range of gradation values 16 to 96) near the least significant bit of the 12-bit output image signal Db is shown. Note that the horizontal axis of FIGS. 3A to 3C indicates the position x of the pixels arranged in the horizontal direction of the screen. Also, the vertical axis in FIG. 3A indicates the gradation value La (x) of the 8-bit input image signal Da at each pixel position, and the vertical axis in FIG. 3B indicates the gradation change amount df. In FIG. 3C, the vertical axis indicates the gradation value Lb (x) of the 12-bit output image signal Db at each pixel position.

In FIG. 3A, the gradation value La (x) of the input image signal Da corresponds to the least significant 1 bit of the gradation value every 8 pixels between the horizontal pixel positions x = 8-40. Although it gradually increases by one gradation, a case where a noise component is included is shown. FIG. 3A shows an example in which a noise component is added to the input image signal that gradually increases by one gradation shown in FIG. In the example of FIG. 3A, the gradation value is increased by one gradation in the pixel at the pixel position x = 19 (= x ₁ ). In the example of FIG. 3A, the gradation value changes in the pixel at the pixel position x = 35 (= x ₂ ), and the pixel at the pixel position x = 34 (= x ₂ −1) before and after the pixel value. In the pixel at the pixel position x = 36 (= x ₂ +1), there is a high-frequency change that increases or decreases the gradation value within the range of one gradation.

In FIG. 3B, the gradation change amount df of the input image signal Da obtained by the gradation conversion means 1, a gradation flat region (df = 0) where there is no change in gradation value, and a noise region (x = x ₁ −1, x ₁ , x ₂ −1, x ₂ , x ₂ +1) and non-noise gradation change positions (pixel positions x = 8, 16, 24, 32, 40) are shown. In FIG. 3 (b), pixels whose pixel values change with a difference of one gradation every eight pixels between horizontal pixel positions x = 8 to 40 (pixel positions x = 8, 16, 24). , 32, 40) is obtained as one gradation (df = + 1). In FIG. 3B, the gradation change amount is also obtained at the pixel position x ₁ and the subsequent pixel position x ₁ +1, and at the pixel position x ₂ and the pixel positions x ₂ −1 and x ₂ +1 before and after the pixel position x _2. df is obtained as a non-zero value. If the change in the gradation value is detected only by the value of the gradation change amount df, the pixel position of the gradation change that should be detected is considered even for the gradation change that has a high frequency change. (For example, a pixel in a noise region is erroneously detected as a pixel in a non-noise gradation change position), which causes false detection.

In general, the human eye is more sensitive to low frequency components than high frequency components, and when quantization noise or pseudo contours are generated and recognized, the gradation difference of change is close to one gradation, to some extent. Is a gradual (low frequency) gradation change in which the gradation value monotonously increases or decreases within a range of. On the other hand, the noise (error) component is a high-frequency change in which the gradation value increases or decreases in a range of several pixels, and its amplitude is considered to be somewhat smaller than the level of the image signal. Thus, variation in gray values of the pixels before and after each of the pixel positions x ₁ and x ₂ in FIG. 3 (b), the increase in area in the range of a few pixels, the change in the reduction frequency, of the change The gradation difference has a relatively small amplitude, and it can be said that these changes are caused by noise components such as noise and error. In order to display the smooth gradation change of the image signal more smoothly in the display image, the flatness / noise area width of the image area where the gradation value changes gently is determined appropriately, excluding the gradation change due to noise components. Therefore, it is necessary to appropriately obtain the gradient of the gradation value so that the gradation value of the pixel of the image signal whose gradation value gradually changes (at a low frequency) changes more smoothly.

  FIG. 3C shows a change in the gradation value Lb (x) of the 12-bit output image signal Db output from the gradation converting means 1. FIG. 3 (c) shows the signal of FIG. 3 (a), the flatness / noise area width of the flat / noise area, which is the area where the gradation value gradually changes (although there is a high-frequency change due to the noise component), An example of the gradation value of the 12-bit output image signal Db output from the gradation converting means 1 when gradation conversion that increases the number of gradations by obtaining the gradient of the gradation value is shown. As shown in the example of FIG. 3C, in the gradation converting unit 1, other gradual gradation changes leave the gradation values smoother while leaving the gradation value changes due to noise components. Gradation conversion is performed so as to change.

<< 1-2-2 >> Tone change detection means
An input image signal Da is input to the gradation change amount detection means 10. The gradation change amount detection means 10 calculates a difference in gradation value (pixel value or gradation level) between pixels adjacent in the horizontal direction in the input image signal Da to obtain a gradation change amount df, which is output. At the same time, an amplitude limited gradation change amount Xlm in which the value of the gradation change amount df is limited within a predetermined range is output.

When the input image signal Da as shown in FIG. 2A or FIG. 3A is input to the gradation change amount detection means 10, the gradation change amount detection means 10 detects the horizontal pixel position x. The difference in gradation value (gradation change amount) between the current pixel and the pixel at the immediately preceding pixel position x−1 is calculated, for example, as shown in FIG. 2B or FIG. A gradation change amount df is obtained. A pixel at a pixel position where the gradation change amount df obtained by the gradation change amount detection means 10 is df ≠ 0 is a pixel at a pixel position where the gradation value changes (that is, a non-noise gradation change position), Alternatively, it is a pixel at a pixel position (that is, a noise region) where the gradation value changes due to a noise component. For example, in the example shown in FIG. 2B, the gradation change amount df is +1 at each pixel position from the pixel position x = 8, and the number of 8 pixels from the pixel position x = 8. In the pixels between the respective pixel positions, the gradation change amount df = 0. On the other hand, in the example shown in FIG. 3B, the gradation change amount df = + 1 from the pixel position x = 8 to the pixel position every 8 pixels, and the pixel position x ₁ (= 19) having the noise component. Is the gradation change amount df = + 1, immediately after the pixel position x ₁ +1 (= 20) is the gradation change amount df = −1, and the pixel position x ₂ (= 35) is the gradation change amount. an amount df = -2, a tone variation df = + 1 in the pixel position _x 2 -1 of the immediately preceding pixel position _{x 2,} gradation change in the pixel position _x 2 +1 immediately after the pixel position _{x 2} The amount df = + 1, and the gradation change amount df = 0 in the other pixels.

  As shown in FIG. 1, the gradation change amount detection means 10 includes, for example, a gradation change amount calculation means 11 and a gradation change amount restriction means 12. The gradation change amount calculation means 11 calculates, for example, a gradation change amount df between pixels adjacent in the horizontal direction from the input image signal Da, and the calculated gradation change amount df The data is output to the means 12, the pixel area determination means 20, and the edge change detection means 60. The gradation change amount limiting unit 12 outputs the amplitude limited gradation change amount Xlm generated based on the gradation change amount df to the inclination detection unit 40 and the correction bit generation unit 50.

When the input image signal Da is input to the gradation change amount calculation unit 11, the gradation change amount calculation unit 11 calculates a difference in gradation values between pixels adjacent in the horizontal direction in the input image signal Da. It is obtained as a tone change amount df. Here, assuming that the horizontal pixel position of the target pixel to be processed is x and the gradation value at this pixel position is La (x), the gradation change amount calculating means 11 is the pixel immediately before the target pixel at the pixel position x. When the gradation value obtained by delaying (the pixel at the pixel position x-1) by one pixel is La (x-1), the gradation value La (x) at the pixel position x is changed to the pixel value at the pixel position x-1. The gradation change amount df is obtained by calculating a value obtained by subtracting the gradation value La (x−1). That is, according to FIG. 2 (a) or FIG. 3 (a), the gradation change amount df, which is the difference in gradation value between pixels, is
df = La (x) −La (x−1)
Can be calculated with When the gradation change amount df, which is a difference in gradation value between pixels, is 0, the gradation is a pixel in a gradation flat region where there is no change in gradation value. In a pixel with a change in value, the absolute value of the difference is the difference in the gradation value, and the direction of the increase or decrease in the gradation value is the sign (polarity) of the value (or a positive sign when the gradation is increased) “+”, And in the case of decrease, it is obtained as a minus sign “−”). The gradation change amount df obtained by the gradation change amount calculation means 11 is sent to the pixel area determination means 20, the edge change detection means 60, and the gradation change amount restriction means 12 in the gradation change amount detection means 10. .

  When the gradation change amount df from the gradation change amount calculation means 11 is input to the gradation change amount restriction means 12, the gradation change amount restriction means 12 causes the gradation change amount df to fall within a predetermined gradation value range. For example, the value of the gradation change amount df is limited so as to be a value (integer) within a range from −1 to +1, and the gradation change amount with the limited value is set as the amplitude limited gradation change amount Xlm. Output as. The range from −1 to +1, which is the limit range here, is the difference of the least significant 1 bit of the 8-bit input image signal Da in the gradation converting means 1. Therefore, the amplitude limit gradation change amount Xlm output from the gradation change amount limiting means 12 is composed of three values of −1, 0, and +1. This amplitude-limited gradation change amount Xlm is output to the inclination detecting means 40 and the correction bit generating means 50, and is used as the height of the gradation change when calculating the inclination of the gradation value with respect to the horizontal direction.

<< 1-2-3 >> Pixel Area Determination Unit
In FIG. 1, the gradation change amount df from the gradation change amount calculation unit 11 is input to the pixel region determination unit 20. The pixel area determination unit 20 determines a change in gradation value due to a noise component from the input gradation change amount df, and determines each pixel as a pixel belonging to the gradation flat area (Ar = 0), noise A determination is made to classify the pixel into either a pixel belonging to the region (Ar = 1) or a pixel at the non-noise gradation change position (Ar = 2), and a region signal Ar indicating the pixel region determination result is generated for each pixel. Output. The area signal Ar is output to the flat / noise area width determining means 30, the inclination detecting means 40, and the correction bit generating means 50.

  As described with reference to FIGS. 2A to 2C and FIGS. 3A to 3C, a gradation change in which the absolute value of the gradation change amount df is one gradation or more or a gradation change due to a noise component If there is, the magnitude of the absolute value of the gradation change amount df becomes a value larger than 0 (| df |> 0), and the gradation change amount df = 0 in a flat pixel without changing the gradation value. It becomes. The noise component is a high-frequency change in which the gradation value increases or decreases in a region of several pixels. The gradation difference of the change has a relatively small amplitude, that is, a relatively small predetermined threshold value Nd (for example, Since the absolute value | df |> 0 of the gradation change amount df, the change in the pixel due to the noise component changes at the pixel position x. The determination can be made based on the value of the gradation change amount df and the polarity (sign indicating positive / negative) of each of the target pixel, the adjacent pixel at the pixel position x−1 immediately before it, and the adjacent pixel at the pixel position x + 1 immediately after the target pixel. In other words, whether or not a gradation change occurs in a pixel due to a noise component among pixels having an absolute value | df |> 0 of the gradation change amount df (whether or not the pixel belongs to a noise region) is determined by the pixel position x. The gradation change amount df (x) of the pixel at the pixel position x1, the gradation change amount df (x-1) of the pixel at the pixel position x-1, and the gradation change amount df (x + 1) of the pixel at the pixel position x + 1 are obtained. The magnitude of the gradation change amount df (x) of the pixel, the polarity of the gradation change amount df (x), the polarity of the gradation change amount df (x−1), and the gradation change amount df (x + 1) It can be determined by the relationship with the polarity.

  4A and 4B are diagrams illustrating an example of a pixel that is determined to be a noise component. FIG. 4A is a graph in which the gradation value is increased by 1 compared to the immediately preceding and immediately following pixels. FIG. 4B shows a noise component consisting of one pixel (one “one gradation changing pixel”), and FIG. 4B shows one pixel whose gradation value is reduced by one gradation compared to the immediately preceding and immediately following pixels. The noise component which consists of a pixel (one "1 gradation change pixel") is shown.

4A and 4B show that the gradation value of one “one gradation changing pixel” is increased by one gradation relative to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. A decrease in gradation (change in high-frequency component) is shown, and an example of a change in the pixel position and gradation value La of a pixel that becomes a noise component is shown. 4A and 4B, the horizontal axis indicates the pixel position x in the horizontal direction, and the vertical axis indicates the gradation value La (x) of the input image signal Da at each pixel position. 4 (a) is increased by one gradation in the pixel position _{x 1,} a noise component decreases 1 gradation pixel position _x 1 +1, 4 (b) is decreased by gradation pixel position _{x 2,} A noise component increasing by one gradation at the pixel position x ₂ +1 is shown.

Figure 4 In the case of (a), the gradation variation is input to the gradation change amount calculation unit 11 is output from the pixel area determination unit 20 df the gray scale variation df _(x 1) at the pixel position _{x 1} = + 1 The gradation change amount df (x ₁ +1) = − 1 at the pixel position x ₁ +1 immediately after that. In the case of FIG. 4B, the gradation change amount df input from the gradation change amount calculation means 11 to the pixel area determination means 20 is the gradation change amount df (x ₂ ) = − 1 at the pixel position x ₂ . At the pixel position x ₂ +1 immediately after that, the gradation change amount df (x ₂ +1) = + 1. As described above, when there is a high-frequency change in the gradation value due to the noise component, the absolute value of the gradation change amount df (x) is less than or equal to a relatively small value (a preset threshold value Nd), and the pixel of interest The sign (polarity) of the gradation change amount of the target pixel with respect to the pixel immediately before the pixel and the sign (polarity) of the gradation change amount of the pixel located immediately after the target pixel with respect to the target pixel are different signs (that is, opposite polarities). There is a change that becomes.

  Therefore, from the magnitude and polarity of the gradation change amount of the target pixel with respect to the pixel immediately before the target pixel, and the magnitude and polarity of the gradation change amount of the pixel located immediately after the target pixel with respect to the target pixel, FIG. Pixels (noise region (Ar = 1) pixels) that are noise components as shown in a) and (b) can be determined. Further, from the gradation change amount df = 0, it is possible to determine the pixels in the gradation flat region (Ar = 0). Further, a pixel that does not belong to any of the pixel in the noise region and the pixel in the flat gradation region can be determined as a pixel at the non-noise gradation change position (Ar = 2). In this way, by determining that each pixel is classified into one of the pixel belonging to the flat gradation region, the pixel belonging to the noise region, and the pixel at the non-noise gradation change position, the region to which each pixel belongs It is possible to generate and output an area signal Ar indicating whether the area is a tonal flat area, a noise area, or a non-noise gradation change position.

  The pixel area determination unit 20 first delays the pixel with respect to the gradation change amount df from the gradation change amount detection unit 10, so that the pixel of interest at the pixel position x and the adjacent pixel at the pixel position x−1 immediately before it are detected. The gradation change amount df in each of the adjacent pixels at the pixel position x + 1 immediately after the target pixel is obtained, and whether or not the change is due to the noise component from the relationship between the magnitude of the obtained gradation change amount df and its polarity. And determining that the pixel at the pixel position x is classified into one of the pixel belonging to the flat gradation region, the pixel belonging to the noise region, and the pixel at the non-noise gradation change position.

  The gradation change amount df at each pixel position has a value of 0 for pixels in a flat gradation area where there is no change in gradation value. Further, the gradation change amount df at the pixel position where the gradation value has changed has the absolute value | df | as the gradation difference, and the change direction (increase or decrease) of the gradation value is the polarity of the gradation change amount df. (I.e., a positive or negative sign). Here, when the gradation value increases, the gradation change amount is positive (df> 0), the polarity is positive (+), and when the gradation value decreases, the gradation change amount is negative (df <0). 0), and its polarity is negative (-). Then, as shown in FIGS. 4A and 4B, the gradation value of one “1 gradation changing pixel” is equal to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. In the case of a noise component in which there is a gradation change that increases or decreases by one gradation, the absolute value | df | of the gradation change amount df is less than or equal to a relatively small preset threshold value Nd, and the pixel of interest The polarity of the gradation change amount (positive or negative sign) is opposite to the polarity of the gradation change amount (positive or negative sign) of any pixel immediately before or after the target pixel.

  Therefore, the gradation change amount df (x) which is the gradation change amount df in each of the target pixel at the pixel position x, the adjacent pixel at the pixel position x−1 immediately before it, and the adjacent pixel at the pixel position x + 1 immediately after the target pixel. ), Df (x−1), df (x + 1), and the following three determinations (that is, determinations A1, A2, and A3), the pixel region in the target pixel at the pixel position x is flattened. It is determined whether the region is a noise region or a non-noise gradation change position.

(Decision A1)
When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0, the target pixel is a pixel belonging to the flat gradation region, and the region signal Ar is Ar = 0.

(Decision A2)
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( x-1) the opposite polarity (when the adjacent pixel has a high frequency component), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( If the polarity is opposite to the polarity of (x + 1) (when the adjacent pixel has a high frequency component),
The target pixel is a pixel belonging to the noise region, and the region signal Ar is Ar = 1.

(Decision A3)
The pixel of interest x at which the pixel of interest x determined that the pixel of interest is not a pixel belonging to the flat gradation region in the determination A1 and that the pixel of interest is not a pixel belonging to the noise region is determined in the determination A2. The pixel at the gradation change position, and the area signal Ar is Ar = 2.

  FIG. 5 is a block diagram schematically showing an example of the configuration of the pixel region determination means 20 shown in FIG. The pixel area determination unit 20 can perform the determinations A1 to A3. FIGS. 6A to 6G are diagrams for explaining an example of the operation of the pixel region determination unit 20 shown in FIGS. 1 and 5, and the input image signal Da including the noise component (FIG. 3). It is a figure for demonstrating said determination A1-A3 with respect to the same as the waveform of (a).

  As illustrated in FIG. 5, the pixel area determination unit 20 extracts, for example, a pixel change amount extraction unit 21 that extracts a gradation change amount of each pixel (that is, the target pixel and its adjacent pixels) used for determining the pixel area. The absolute value level comparing means 22 for comparing the magnitude (level) of the absolute value | df | of the gradation change amount df with a predetermined threshold, the adjacent position polarity determining means 23 and 24, the synthesizing means 25, the area Signal generating means 26. Although the adjacent position polarity determination means 23 and 24 have different inputs, both can have the same configuration.

  In FIG. 5, the gradation change amount df input to the pixel area determination unit 20 is input to the pixel change amount extraction unit 21. The pixel change amount extraction unit 21 includes, for example, two pixel delay units (indicated by “D” in FIG. 5) 211 and 212. The pixel change amount extraction unit 21 delays the gradation change amount df by a pixel, and changes the gradation change amount df (x) in the target pixel at the pixel position x and adjacent pixel positions located before and after the target pixel at the pixel position x. The gradation change amounts df (x−1) and df (x + 1) at x−1 and x + 1 are extracted. The determinations A1 to A3 for classifying into three pixel areas based on the change in the pixel and the gradation value at the pixel position x based on the gradation change amounts df (x), df (x-1), and df (x + 1) are performed. .

  The absolute value level comparison means 22 receives the gradation change amount df (x) at the target pixel at the pixel position x. The absolute value level comparison means 22 compares the absolute value level of the gradation change amount df with a predetermined threshold value. The predetermined threshold value is, for example, a value 0 and a relatively small value Nd corresponding to the gradation difference in the noise component (for example, a value twice the minimum gradation difference, Nd = 2). When the value of the gradation change amount df (x) is 0 (when the target pixel is determined to be a pixel in the flat gradation region (df (x) = 0) in the determination A1), and the gradation change The absolute value level of the gradation change amount df (x) is determined by determining that the absolute value of the amount | df (x) | ≦ Nd (when one of the conditions in the determination A2 is satisfied). The determination result is output as the comparison result nlf.

  The value of the comparison result nlf obtained by the absolute value level comparing means 22 is, for example, when the gradation change amount df (x) = 0, the comparison result nlf = 1 and the gradation change amount df (x) ≠ 0. If the absolute value of the gradation change amount | df (x) | ≦ Nd, the comparison result nlf = 2, and otherwise (Nd <| df (x) |), the comparison result nlf = 0. Note that the value of the comparison result nlf is not limited to the above value, and may be any other value as long as the comparison result of the absolute value level of the gradation change amount df is indicated. The predetermined threshold Nd will be described as Nd = 2. However, the predetermined threshold Nd is sufficiently smaller than the image signal, and the amplitude of noise components such as noise and error included in the input image signal is large. May be set to another value as long as it is a value that takes into account (a value larger than the amount of gradation change that is desired to be determined as a noise component).

  The adjacent position polarity determination unit 23 receives the gradation change amount df (x) obtained by the pixel change amount extraction unit 21 and the gradation change amount df (x−1). The adjacent position polarity determination unit 24 receives the gradation change amount df (x) and the gradation change amount df (x + 1) obtained by the pixel change amount extraction unit 21. The adjacent position polarity determination means 23 and 24 differ in that the gradation change amount df (x−1) is input to the former and the gradation change amount df (x + 1) is input to the latter, but the configuration is the same. It is. The adjacent position polarity determination means 23 and 24 determine whether or not the pixel of interest at the pixel position x satisfies one of the conditions in the determination A2 (condition that adjacent pixels in the noise region have a high-frequency component change). .

  The adjacent position polarity determination unit 23 performs the opposite direction at the pixel position x-1 adjacent to the pixel position x based on the polarity of the gradation change amount df (x) and the polarity of the gradation change amount df (x-1). Whether or not the polarity of the gradation change amount df (x) and the gradation change amount df (x-1) are opposite (signs do not match). As a determination result at the position x, a determination result pf1 indicating a binarized value is obtained. The value of the determination result pf1 is, for example, the value “1” (pf1 = 1) when the polarity of the gradation change amount is reversed, and the value “0” (pf1 = 0) when the polarity is the same. Note that the value of the determination result pf1 is not limited to the above value, and may be other values as long as the signs indicate mismatch or coincidence.

  Similarly, the adjacent position polarity determination unit 24 performs a change in the opposite direction at the pixel position x + 1 adjacent to the pixel position x based on the polarities of the gradation change amount df (x) and the gradation change amount df (x + 1). Whether or not the polarity of the gradation change amount df (x) and the gradation change amount df (x + 1) are opposite (signs do not match), and the determination result at the pixel position x As a result, a determination result pf2 indicating a binarized value is obtained. The value of the determination result pf2 is, for example, the value “1” (pf2 = 1) when the polarity of the gradation change amount is reversed, and the value “0” (pf2 = 0) when the polarity is the same. Note that the value of the determination result pf2 is not limited to the above value, and may be another value as long as the sign indicates a mismatch or a match.

  Determination results pf1 and pf2 from the adjacent position polarity determination units 23 and 24 are input to the combining unit 25. The synthesizing unit 25 synthesizes the determination results pf1 and pf2, and the polarity determination result when the polarity of the gradation change amount of any of the adjacent pixels is opposite to the gradation change amount df (x) at the pixel position x. Find pof. For example, when the determination result pf1 = 1 or the determination result pf2 = 1, the synthesizing unit 25 outputs the polarity determination result pof = 1. As a result, it is possible to obtain a result of determining a case where the gradation change amount df (x) at the pixel position x is opposite in polarity to the gradation change amount in any of the adjacent pixels immediately before or after the pixel position x. .

  The region signal generation means 26 receives the comparison result nlf from the absolute value level comparison means 22 and the polarity determination result pof from the synthesis means 25. The area signal generation unit 26 classifies the pixel areas according to the determinations indicated by the determinations A1 to A3, and generates and outputs an area signal Ar indicating each area determination result. That is, the area signal generation unit 26 performs the determinations indicated by the determinations A1 to A3, and assigns each pixel to any one of a pixel belonging to the gradation flat area, a pixel belonging to the noise area, and a pixel at the non-noise gradation change position. A determination to classify is performed, and an area signal Ar having a value corresponding to the determination result is generated. As a result, an area signal Ar for classifying the change in gradation value in each pixel into three areas is obtained, and each pixel is in any area of the gradation flat area, the noise area, or the non-noise gradation change position. Indicates whether it belongs.

  6A shows the pixel position x in the horizontal direction of the input image signal Da, FIG. 6B shows the gradation value La (x) of the input image signal Da, and FIG. FIG. 6 (d) to (f) are obtained by the pixel area determination unit 20 of FIG. 5 and FIG. 6 (d) to (f) are shown. FIG. 6G shows the comparison result nlf of the level of the gradation change amount df (x) and the determination results pf1 and pf2 of the polarity determination with the adjacent pixels. FIG. Indicates.

For example, in FIGS. 6A to 6G, the gradation change amount df (f) due to the difference in gradation values between adjacent pixels obtained for the input image signal Da including the noise component (FIG. 6B). FIG. 6C is obtained as the gradation change amount df = + 1 at the pixel position x = 16, 24, 32, 40, but the pixel position x ₁ and the subsequent pixel position x ₁ +1 and the pixel position x ₂ and the pixel noise components at pixel positions x ₂ −1 and x ₂ +1 before and after that, the gradation change amount df is obtained as a non-zero value. When the gradation change amount is input to the pixel area determination unit 20, a gradation change amount _{df (x 1) = + 1} ≦ Nd at a pixel position _{x 1,} gradation change at the pixel position _x 1 +1 to the adjacent Since the amount df (x ₁ +1) = − 1 <0, the comparison result nlf = 2 from the absolute value level comparison unit 22 in FIG. 5 and the determination result pf2 = 1 from the adjacent position polarity determination unit 24, that is, The polarity determination result pof = 1 from the combining unit 25 is obtained. Therefore, the area signal Ar output from the area signal generating means 26 has a value “1” indicating the noise area.

When the above-described determinations A1 to A2 are performed on the other pixels in the pixel area determination unit 20, the pixel at the pixel position x ₁ +1, the pixel at the pixel position x ₂ , the pixel at the pixel position x ₂ -1, and the pixel position x ₂ +1 In this pixel, the area signal Ar = 1. Thus, in the example shown in FIG. 6 (a) ~ (g) , the pixels of the pixel and the pixel position x ₁ +1 pixel position x _1, it is determined that the pixel of the noise region. In addition, the pixel at the pixel position x ₂ −1, the pixel at the pixel position x ₂ , and the pixel at the pixel position x ₁ +1 are determined to be pixels in the noise region.

  As described above, the pixel area determination unit 20 determines the value of the gradation change amount df (x) and the gradation change amount df from the gradation change amounts df (x), df (x−1), and df (x + 1). The polarity of (x), the gradation change amount df (x-1), and the gradation change amount df (x + 1) are obtained, and the above determinations A1 to A3 are used to determine the pixel of interest at the pixel position x. It is possible to determine whether the pixel belongs to any one of the gradation flat area, the noise area, and the non-noise gradation change position, and obtain an area signal Ar indicating the area determination result of the pixel. This determination method does not determine the pixel region by detecting the gradation difference within the same gradation value or within a certain value, but determines the gradation change due to the noise component by the magnitude of the value of the gradation change amount df, Since it is determined by the polarity, the influence of noise is small, and it is a pixel that belongs to any one of the gradation flat area, noise area, and non-noise gradation change position more accurately by only the gradation change amount in the adjacent pixel. Can be determined.

  5 is configured such that the pixel change amount extraction unit 21 extracts the gradation change amounts df (x), df (x−1), and df (x + 1). When obtaining the gradation change amount df in the change amount detection means 10, it may be obtained by obtaining the gradation change amount at the pixel position x and its adjacent pixels. In this case, the pixel in the pixel change amount extraction means 21 is obtained. Delay can be omitted. Each of the adjacent position polarity determination means 23 and 24 performs determination based on the polarity of the gradation change amount df (x) and the polarity of the gradation change amount df (x−1), and the gradation change amount. Since the determination is made based on the polarity of df (x) and the polarity of the gradation change amount df (x + 1), the adjacent pixel position x−1 of the pixel position x obtained by the pixel change amount extraction means 21 or the gradation change amount detection means 10. As for the tone change at x + 1, only the polarity may be obtained.

<< 1-2-4 >> Flat / Noise Area Width Determination Unit 30.
The flat area / noise area width determination means 30 receives the area signal Ar from the pixel area determination means 20 and a synchronization signal (not shown) (for example, a signal synchronized with the horizontal scanning period). The flat / noise area width determining means 30 has a width of a flat period (which may include a noise component) with no gradation change based on the area signal Ar from the pixel area determining means 20 for each horizontal scanning period. Is a flat / noise area width (if no noise component is included, it is the width of the gradation flat area). In the first embodiment, the flat / noise area width determination unit 30 uses, for example, the non-noise gradation change position indicated by the area signal Ar = 2 as a reference pixel, and performs flatness up to the pixel immediately before the non-noise gradation change position. A noise area width is obtained, and a flat / noise area width data predicted value dst that is a predicted value of the flat / noise area width after the non-noise gradation change position is obtained from the flat / noise area width. In the first embodiment, a memory for storing pixel data of the number of pixels corresponding to the detected flat / noise area width is not required.

  Specifically, the flat / noise area width determining means 30 initializes every horizontal scanning period (outputs dst = 0), and based on the area signal Ar, the non-noise level indicated by the area signal Ar = 2. Using the tone change position as the reference pixel, the number of pixels in the flat / noise area up to the pixel immediately before the reference pixel was obtained, and the flat / noise area width after that was predicted for each non-noise gradation change position (Ar = 2). A flat / noise area width data prediction value dst to be a value is obtained. The obtained flat / noise area width data predicted value dst is output to the inclination detecting means 40, and the gradient width (distance) of the gradation value change is calculated when calculating the gradient of the gradation value change in the horizontal direction. ). Since the number of pixels is counted based on the non-noise gradation change position indicated by the area signal Ar = 2, the pixel of the flat gradation area where the area signal Ar = 0 and the pixel of the noise area indicated by the area signal Ar = 1 Both pixels are counted as pixels in the flat / noise region. Therefore, the flat / noise area width determination unit 30 can obtain the number of pixels between the non-noise gradation change positions (Ar = 2) including the noise area (that is, without erroneous detection due to the noise component).

  FIG. 7 is a block diagram schematically showing an example of the configuration of the flat / noise area width determining means 30 shown in FIG. The flat / noise area width determining means 30 is the number of pixels in a flat period with no change in gradation value in the horizontal direction (that is, the total number of pixels in the gradation flat area and the number of pixels in the noise area. For each non-noise gradation change position, a flat / noise area width data predicted value dst obtained by predicting the flat / noise area width after the non-noise gradation change position is obtained. As shown in FIG. 7, the flat / noise region width determining unit 30 includes, for example, a flat / noise region width counting unit 31 and a cyclic filter processing unit 32. The cyclic filter processing unit 32 includes, for example, a coefficient 1 multiplication unit 33, a calculation unit 34, a coefficient 2 multiplication unit 35, a switching unit 36, and a filter output delay unit 37.

  An area signal Ar from the pixel area determination unit 20 is input to the flat / noise area width counting unit 31. The flat / noise area width counting means 31 initializes the count value for each horizontal scanning period (sets the count value to 0), and uses the non-noise gradation change position indicated by the area signal Ar = 2 as a reference. The flat / noise area width in the horizontal direction up to the previous pixel is obtained and output as a flat / noise area width count value Cnt. That is, the flat / noise area width counting means 31 initializes the pixel count value to 0 at the non-noise gradation change position (Ar = 2), and the gradation flat area and noise area where the area signal Ar = 0, 1 is obtained. And the count value of the number of pixels counted up to the immediately preceding pixel at the non-noise gradation change position (Ar = 2) is calculated as a flat / noise area width count value Cnt indicating a flat / noise area width. Output as.

For example, when the area signal Ar shown in FIG. 6G is input to the flat / noise area width counting means 31, the pixel position where the area signal Ar = 2 is the non-noise gradation change position (pixel). The number of pixels between the pixels at the position x = 16, 24, 32, 40) and the region signal Ar = 2 is counted, and the number of pixels up to the pixel immediately before the non-noise gradation change position is flat / noise region. The flat / noise area width count value Cnt indicating the width is obtained. In the example of FIG. 6G, the pixel position x ₁ (= 19), the pixel position x ₁ +1 (= 20), the pixel position x ₂ (= 35), and the pixel position x ₂ −1 (= 34). ) And the pixel position x ₂ +1 (= 36) are determined as noise areas, are included in the flat / noise area width count, and are not erroneously detected as non-noise gradation change positions (Ar = 2). .

  The flat / noise area width count means 31 obtains the flat / noise area width in the horizontal direction up to the previous pixel at the non-noise gradation change position (Ar = 2) as the flat / noise area width count value Cnt. The flat / noise area width count value Cnt is output to the recursive filter processing means 32.

  The recursive filter processing means 32 receives the area signal Ar from the pixel area determining means 20, and the recursive filter processing means 32 further receives the flat / noise area width from the flat / noise area width counting means 31. A count value Cnt is input. In the cyclic filter processing means 32, the output result is initialized for each horizontal scanning period, and the non-noise gradation change with respect to the flat / noise area width count value Cnt at the non-noise gradation change position (Ar = 2). Flat / noise area width data prediction, which is a value obtained by performing a cyclic filter process at each position (Ar = 2), and predicting the flat / noise area width after the pixel position where the gradation value changes based on the result of this filter process. Output as the value dst. The recursive filter processing means 32 is a low-pass filter (LPF), and a recursive filter (Infinite Impulse Response (IIR)) based on the input signal and the filter output pdst delayed by the filter output delay means 37 in the recursive filter processing means 32. ) Filter). The configuration by the IIR filter reduces the number of delay means (memory) to be used as compared with the configuration in the acyclic filter (Finite Impulse Response (FIR) filter) using signals of several adjacent pixels of the input signal, and the filter processing It can be constant regardless of the pixel range to be performed.

In FIG. 7, the coefficient 1 multiplication means 33 and the coefficient 2 multiplication means 35 are set with predetermined filter coefficients, and perform an operation of multiplying each input signal by the filter coefficient. Here, when the filter coefficient in the coefficient 1 multiplication means 33 is k ₁ and the filter coefficient of the coefficient 2 multiplication means 35 is k ₂ , k ₁ = 1−k _2, 0 ≦ k ₁ ≦ 1, 0 ≦ k ₂ ≦ The value of the filter coefficients k ₁ and k ₂ is set to 1 and the size of the past output value to be circulated, that is, the range of the region affected by the obtained flat / noise region width is determined.

The coefficient 1 multiplication means 33 receives the flat / noise area width count value Cnt from the flat / noise area width count means 31, multiplies the flat / noise area width count value Cnt by the filter coefficient k ₁ , and the multiplication result k ₁ × Get Cnt. The coefficient 2 multiplication means 35 is a delayed flat / noise area width data predicted value pdst (that is, before the current pixel position), which is a filter processing output at a pixel position immediately before being delayed by one pixel by a filter output delay means 37 described later. The result after filtering at the non-noise gradation change position) is multiplied by the delay flat / noise region width data prediction value pdst by the filter coefficient k ₂ to obtain a multiplication result k ₂ × pdst.

The calculation means 34 calculates the coefficient multiplication result k ₁ × Cnt for the flat / noise area width count value Cnt output from the coefficient 1 multiplication means 33 and the delayed flat / noise area width data predicted value output from the coefficient 2 multiplication means 35. The coefficient multiplication result k ₂ × pdst for pdst is input. The computing unit 34 adds the input coefficient multiplication results and outputs the addition result as a filter processing result fdst. Since the delay flat / noise region width data prediction value pdst is the filter processing output at the pixel position immediately before being delayed by one pixel by the filter output delay unit 37, the coefficient 1 multiplication unit 33, the coefficient 2 multiplication unit 35, and the calculation unit The calculation by 35 corresponds to cyclic filter processing and is calculated by the following equation (1).
fdst
= K ₂ × pdst + k ₁ × Cnt
= K ₂ × pdst + (1−k ₂ ) × Cnt Equation (1)

  Further, the calculation of the cyclic filter processing according to the above expression (1) represents the calculation by the low-pass filter, and the low-frequency component of the flat / noise area width when the gradation value changes in the horizontal direction, That is, smoothed continuity is obtained, and it is possible to avoid detecting a sudden change in gradation value as a flat / noise area width.

  The switching means 36 receives the area signal Ar, the filter processing addition result fdst from the calculation means 35, and the delay flat / noise area width data predicted value pdst from the filter output delay means 37. The switching means 36 switches between the filter processing result fdst from the computing means 34 and the delay flat / noise area width data predicted value pdst from the filter output delay means 37 according to the pixel area indicated by the area signal Ar, and also initializes it. In this case, the value is switched to 0 and output, and is output as a predicted flat / noise region width data value dst, which is a predicted value of the flat / noise region width after the non-noise gradation change position (Ar = 2).

  In other words, the switching means 36 selects the initialization value 0 and outputs it as the flat / noise area width data predicted value dst = 0 when initialization is indicated at the interval of the horizontal scanning period. Further, the switching means 36 at the non-noise gradation change position (Ar = 2) has a filter processing result from the arithmetic means 34 that is a result of applying a cyclic filter process to the flat / noise area width count value Cnt. fdst is selected and output as a flat / noise region width data prediction value dst. On the other hand, in the gradation flat region (Ar = 0) and noise region (Ar = 1), the switching unit 36 determines the delay flat / noise region width data predicted value pdst (that is, the current pixel) from the filter output delay unit 37. (Corresponding to the filter processing result fdst at the non-noise gradation change position before the position) is selected and output as the flat / noise area width data predicted value dst.

  As a result, in the recursive filter processing means 32, the result of performing recursive filter processing on the flat / noise region width count value Cnt for each non-noise gradation change position (Ar = 2) is the flat / noise region. It is obtained as the width data predicted value dst, and in the other gradation flat region and noise region pixels, the value at the previous non-noise gradation change position is held and output. The flat / noise region width data predicted value dst from the switching unit 36 is output to the inclination detecting unit 40 as an output from the cyclic filter processing unit 32 and also to the filter output delay unit 37.

  The filter output delay means 37 delays the flat / noise area width data predicted value dst from the switching means 36 by one pixel and outputs the delayed flat / noise area width data predicted value pdst. Since the filter output delay means 37 delays the output from the switching means 36, the delay flat / noise area width data predicted value pdst is the flat / noise area width data predicted value dst of the previous pixel output as it is, The result (cyclic / noise area width data predicted value dst) of the cyclic filter processing for the flat / noise area width count value Cnt at each non-noise gradation change position (Ar = 2) is the next non-noise gradation change position. Up to pixels where the tone value changes.

  As described above, the cyclic filter processing unit 32 performs cyclic filter processing for each pixel position where the gradation value changes with respect to the flat / noise area width count value Cnt at the non-noise gradation change position (Ar = 2). The flat / noise region width data prediction value dst subjected to the processing is output. The flat / noise area width data predicted value dst is obtained by performing a cyclic low-pass filter process on the flat / noise area width count value Cnt at the non-noise gradation change position (Ar = 2). Thus, the low-frequency component of the flat / noise region width when the gradation value changes in the horizontal direction, that is, the flat / noise region width value indicating smoothed continuity is obtained. Therefore, in the pixel area where the gradation value changes gradually and gradually in the horizontal direction, the flat / noise area width when the gradation value changes (the width of the gradation flat area when there is no noise component) ) Can be obtained while avoiding steep changes, and as a result, the flatness / noise area width of the gradation value change after the non-noise gradation change position (Ar = 2) can be used as a predicted value. it can.

<< 1-2-5 >> Edge change detection means 60.
The edge change detection means 60 receives the gradation change amount df from the gradation change amount calculation means 11 in the gradation change amount detection means 10. The edge change detecting unit 60 sets a correction limiting coefficient rc used to limit the correction value when the correction bit generating unit 50 generates a correction value (correction bit data Cd) for expanding the number of gradations. The correction limit coefficient rc is set and output to the correction bit generation means 50.

  Here, the gradation change amount df from the gradation change amount detection means 10 is calculated from the difference in gradation value between adjacent pixels, and in the pixel adjacent to the edge (contour) of the image, It also includes a difference in gradation values due to edge components. In the first embodiment, an area of a pixel whose number of gradations is expanded by the gradation converting unit 1 is an area in which the difference in gradation values is a change corresponding to the least significant 1 bit of the input image signal Da (very smooth). The area in which the gradation value changes). For this reason, it is necessary to prevent a correction value from being added to the gradation value at the edge of the image. Therefore, the edge change detecting means 60 determines whether or not the change in the gradation value (that is, the difference in the gradation value) is due to the edge of the image based on the magnitude of the gradation change amount df. Based on the result, the correction limiting coefficient rc is set as a value that can limit the correction value so that the gradation correction value is not added to the edge of the image.

  FIG. 8 is a block diagram schematically showing an example of the configuration of the edge change detection means 60. In FIG. 8, the edge change detecting means 60 calculates the correction limiting coefficient rc from the magnitude of the absolute value of the gradation change amount df and the magnitude of the low frequency component (low frequency component) of the gradation change amount df. Shows the configuration for setting. The edge change detection means 60 uses the low frequency component of the gradation change amount df, not only the gradation change amount df, so that the gradation due to the edge of the image in the pixel near the pixel position where the gradation value changes. A change in value can also be taken into account, and it can be determined whether or not the pixel has an edge of the image with higher accuracy. As shown in FIG. 8, the edge change detection means 60 includes, for example, an absolute value calculation means 61, a differential low frequency component extraction means 62, an absolute value calculation means 63, a magnification conversion means 64, and a maximum value selection means. 65 and coefficient conversion means 66.

  When the gradation change amount df is input to the absolute value calculation means 61 in the edge change detection means 60, the absolute value calculation means 61 calculates the absolute value of the gradation change amount df, and the gradation change amount is obtained as a calculation result. The absolute value dxc is output.

  When the gradation change amount df from the gradation change amount calculation means 11 is input to the difference low-frequency component extraction means 62, the difference low-frequency component extraction means 62 applies a horizontal LPF (horizontal) to the gradation change amount df. LPF) processing is performed to smooth the gradation change amount df, thereby extracting and outputting a low frequency component (referred to as “differential low frequency component”) dflp of the gradation change amount df. When considering the memory capacity, the differential low-frequency component extracting means 62 can be reduced in circuit scale if constituted by an IIR filter that performs filter processing by circulating the filter output. Further, the difference low-frequency component extracting unit 62 may use a value obtained by adding the gradation change amount df between adjacent pixels.

  When the difference low-frequency component dflp from the difference low-frequency component extraction means 62 is input to the absolute value calculation means 63, the absolute value calculation means 63 calculates the absolute value of the difference low-frequency component dflp, and the difference is low as the calculation result. The band component absolute value abslp is output.

  When the difference low-frequency component absolute value abslp from the absolute value calculation means 63 is input to the magnification conversion means 64, the magnification conversion means 64 adds a predetermined magnification to the difference low-frequency component absolute value abslp from the absolute value calculation means 63. Multiplication is performed to perform gain adjustment (that is, conversion processing of the differential low-frequency component absolute value abslp), and the magnification conversion differential low-frequency component dxlpf obtained by the conversion processing is output to the maximum value selection means 65. The difference low-frequency component absolute value abslp is a value obtained as an absolute value of the result of extracting the low-frequency component of the gradation change amount df in the difference low-frequency component extraction unit 62 (that is, the differential low-frequency component dflp). The gradation value is different from the gradation change amount absolute value dxc from the absolute value calculation means 61 that has not undergone the low-frequency component extraction processing by the differential low-frequency component extraction means 62. The magnification conversion means 64 adjusts the gain of the difference low-frequency component absolute value abslp from the absolute value calculation means 63 and converts it into a magnification conversion difference low-frequency component dxlpf, thereby converting the gradation value of the magnification conversion difference low-frequency component dxlpf. Then, the gradation value is adjusted to a gradation value corresponding to the gradation value of the gradation change amount absolute value dxc from the absolute value calculation means 61. Note that by changing the gain adjustment magnification in the magnification conversion means 64, it is possible to adjust the detection sensitivity of the edge of the image in determining whether the gradation difference is a gradation difference due to the edge of the image. it can.

  When the gradation change amount absolute value dxc from the absolute value calculator 61 and the magnification conversion difference low-frequency component dxlpf from the magnification converter 64 are input to the maximum value selector 65, the maximum value selector 65 is input. The gradation change amount absolute value dxc and the magnification conversion difference low-frequency component dxlpf are compared, the maximum value among these values is selected, and the selected maximum value is output as the maximum difference absolute value | df |. . That is, since the maximum difference absolute value | df | indicates the difference between the edge of the horizontal image or the gradation value in the vicinity of the target pixel, the presence of the edge of the image is determined from the magnitude of the difference value of the gradation value. The presence or absence of can be determined. The maximum difference absolute value | df | from the maximum value selecting means 65 is output to the coefficient converting means 66.

  When the maximum difference absolute value | df | from the maximum value selection unit 65 is input to the coefficient conversion unit 66, the coefficient conversion unit 66 calculates the floor from the value of the maximum difference absolute value | df | from the maximum value selection unit 65. It is determined whether or not the change in the tone value is due to the edge of the image, and based on the result, the generated correction value is limited so that the gradation correction value is not added to the input image signal Da at the edge of the image. The correction limit coefficient rc is set as a signal that can be transmitted, and the set correction limit coefficient rc is output to the correction bit generation means 50.

  Here, the correction limiting coefficient rc set by the edge change detection means 60 is a multiplication value for the gradation correction value, and is set as a value in the range from 0 to 1. Since the correction limiting coefficient rc is a multiplication value for the gradation correction value, if the correction limiting coefficient rc is 0, the correction value is also 0. In addition, in the pixel region where the gradation value changes smoothly and the number of gradations is expanded, the edge change detection means 60 adds correction bit data based on the correction value to the value of the image signal. The correction limiting coefficient rc is set to 1.

The coefficient converting unit 66 calculates the value of the correction limiting coefficient rc according to the value of the maximum difference absolute value | df | from the maximum value selecting unit 65. For example, the coefficient conversion unit 66 compares the maximum difference absolute value | df | with a predetermined value TH, and if the maximum difference absolute value | df | is equal to or less than the predetermined value TH, the correction limiting coefficient rc is set. When the maximum difference absolute value | df | is between the value TH and twice the value TH (2 × TH), the correction limit coefficient rc approaches 0 as the maximum difference absolute value | df | increases. Thus, the value of rc is changed, and for example, it is calculated as a value like the following formula (2).
rc = (2 × TH− | df |) / TH (2)
When the maximum difference absolute value | df | is equal to or greater than 2 × TH, the correction limiting coefficient rc is set to 0. In this case, it is determined that the change in the gradation value is due to the edge of the image.

  The predetermined value TH to be compared is a difference of the least significant 1 bit of the input image signal Da, that is, an absolute value of +1 and −1 which is a difference in gradation value between pixels, and 1 do it.

  With the above configuration, the correction limiting coefficient rc that changes within the range from 0 to 1 according to the value of the maximum difference absolute value | df | is obtained, and the maximum difference absolute value | df | is small (TH = 1 or less). In the pixel region where the gradation value changes smoothly, the correction limiting coefficient rc = 1, the maximum difference absolute value | df | is large, and it is determined that the difference in gradation is due to the edge of the image. The limit coefficient rc is 0 or smaller than 1, and the correction limit coefficient rc can be obtained as a value that can limit the correction value so that the gradation correction value is not added to the edge of the image.

  FIG. 9 is a diagram showing the characteristics of the conversion into the correction limiting coefficient rc (vertical axis) with respect to the maximum difference absolute value | df | (horizontal axis) in the coefficient conversion means 66. FIG. 9 shows a case where the coefficient conversion means 66 outputs the correction limiting coefficient rc as a value within the range from 0 to 1. When the maximum difference absolute value | df | increases, the value of the correction limiting coefficient rc approaches 0, and it is determined that the change in the gradation value is caused by the edge of the image, and the correction value is limited so that the number of gradations is not expanded. On the other hand, when the value of the maximum difference absolute value | df | is equal to or smaller than the threshold value TH = 1, it is assumed that the correction limiting coefficient rc = 1 and that the gradation value is a gradual change in the difference between +1 and −1. It is set to extend the number of gradations. As a result, the correction limit coefficient rc is set to rc = 0 in the pixel determined to have an image edge near the pixel where the gradation value changes, and the correction value is limited by the value.

FIGS. 10A and 10B are diagrams illustrating an example of a change in the gradation value at which the correction limiting coefficient rc set by the edge change detection unit 60 is zero. 10A and 10B, the horizontal axis indicates the pixel position in the horizontal direction, and the vertical axis indicates the gradation value La (x) of the input image signal Da at each pixel position. For example, in the case shown in FIG. 10A, the absolute value of the gradation change amount df at the pixel position x ₀ where the gradation value changes, that is, the maximum difference absolute value | df | is larger than the threshold value TH = 1. since, in 0 the value of the correction limit coefficient rc at the pixel position x _0, variation in gray value is determined to be due to the edge of the image. In the case shown in FIG. 10 (b), the difference low frequency component shown in tone variation df = 3, and the 8 at the pixel position x ₀ +1 adjacent to the pixel position x ₀ of the gradation value changes When dflp is a value based on the horizontal LPF with the left and right pixels, the maximum difference absolute value | df | becomes larger than the threshold value TH = 1, the value of the correction limiting coefficient rc approaches 0, and the change in the gradation value is an image. It is determined to be due to the edge. Then, the pixel position x ₀ and subsequent pixels in FIG. 10 (a) and (b) than the value of the correction limiting coefficient rc, the correction value is limited.

  Although the case where the coefficient conversion unit 66 sets the correction limiting coefficient rc from the comparison with the predetermined value TH and the calculation according to the equation (2) has been described, the ROM (Read) is based on the maximum difference absolute value | df |. The correction limiting coefficient rc may be set by conversion obtained in advance, for example, as shown in FIG. Further, the maximum difference absolute value | df | may be a fixed value within a range from 0 to 1 even for a value between the value TH and a value twice the value TH (2 × TH). Furthermore, the value of the correction limit coefficient rc may be a binary signal of 0 and 1. The correction limit coefficient rc indicates the correction value of a pixel determined to have an image edge near the pixel where the gradation value changes. Any value that can be limited is acceptable.

  As described above, the edge change detection means 60 in FIG. 8 obtains the correction limiting coefficient rc that changes according to the magnitude of the gradation change amount df as a value within the range from 0 to 1, and corrects the correction. Output to the bit generation means 50. Note that the configuration of the edge change detection means 60 is not limited to the illustrated example, and other configurations may be employed.

<< 1-2-6 >> Inclination detecting means 40.
The inclination detection means 40 includes an area signal Ar from the pixel area determination means 20, an amplitude limit gradation change amount Xlm from the gradation change amount restriction means 12, and a recursive filter in the flat / noise area width determination means 30. The flat / noise area width data predicted value dst from the processing means 22 is input. The inclination detecting unit 40 determines the level of the input image signal Da in the horizontal direction from the amplitude limit gradation change amount Xlm and the flat / noise area width data predicted value dst at the pixel position where the gradation value indicated by the area signal Ar changes. The inclination of the tone value is detected, the inclination data value Ksl is obtained, and the obtained inclination data value Ksl is output to the correction bit generation means 50. Since the amplitude limit gradation change amount Xlm is input as a value in which the gradation change amount df is limited within the range from −1 to +1, the gradient of the detected gradation value ranges from −1 to +1. It becomes the value in.

  FIG. 11 is a block diagram schematically showing an example of the configuration of the inclination detecting means 40. As shown in FIG. 11, the inclination detecting means 40 is composed of, for example, an inclination calculating means 40a, a switching means 40b, and an inclination delay means 40c.

In FIG. 11, the slope calculating unit 40 a includes the amplitude limited gradation change amount Xlm from the gradation change amount limiting unit 12 and the flat / noise region width data predicted value dst from the flat / noise region width determining unit 30. Entered. The inclination calculation means 40a calculates a fine inclination divk of the gradation value from the value of the amplitude limit gradation change amount Xlm and the flat / noise area width data predicted value dst by the following equation (3).
divk = Xlm / dst (3)

  The amplitude limit gradation change amount Xlm is limited within a range of −1 to +1. For example, the amplitude limit gradation change amount Xlm is +1, and the flat / noise area width is changed from the change of the gradation value. Is input with a predicted flat / noise area width data value dst = 7, where the calculated slope divk is 1/7. When the amplitude limit gradation change amount Xlm is −1, the slope divk = −1 / 7. When the flat / noise region width data prediction value dst = 0, the slope divk = Xlm. Therefore, the slope divk is obtained as a value within a range from −1 to +1.

  The switching means 40b receives the area signal Ar from the pixel area determination means 20, the inclination divk calculated by the inclination calculation means 40a, and the delay inclination data value pKsl delayed by one pixel from the inclination delay means 40c. The switching means 40b switches between the slope divk calculated by the slope calculation means 40a and the delay slope data value pKsl from the slope delay means 40c in accordance with the area signal Ar from the pixel area determination means 20 to thereby change the horizontal gradation. An inclination data value Ksl that is a result of detecting the inclination of the value is output.

  That is, the switching means 40b selects the inclination divk calculated by the inclination calculation means 40a at the non-noise gradation change position (Ar = 2) based on the area signal Ar from the pixel area determination means 20, and the inclination data value Ksl. And On the other hand, in the gradation flat region (Ar = 0) and the noise region (Ar = 1), the switching unit 40b performs the delay slope data value pKsl (that is, non-noise before the current pixel position) from the slope delay unit 40c. Is selected as the gradient data value Ksl. This inclination data value Ksl is output to the correction bit generation means 50 as an inclination detection result in the inclination detection means 40 and is also sent to the inclination delay means 40c, delayed by one pixel, and switched as a delay inclination data value pKsl by the switching means 40b. Returned to

  The slope delay means 40c delays the slope data value Ksl from the switching means 40b by one pixel and outputs a delayed slope data value pKsl. Since the inclination delay means 40c delays the output from the switching means 40b, the inclination inclination data value pKsl is output as it is as the inclination data value Ksl at the previous pixel, and is calculated for each non-noise gradation change position (Ar = 2). The value of the slope divk is held up to the pixel where the gradation value that becomes the next non-noise gradation change position changes.

  Therefore, the gradient of the gradation value calculated by the inclination calculation means 40a is obtained as the inclination data value Ksl for each pixel at the non-noise gradation change position (Ar = 2) by the switching means 40b and the inclination delay means 40c. In the gradation flat region (Ar = 0) and noise region (Ar = 1), the value of the pixel that is the non-noise gradation change position (Ar = 2) immediately before the current pixel is held and output. The

  For example, assuming that the amplitude limit gradation change amount Xlm at the non-noise gradation change position (Ar = 2) is +1 and the flat / noise area width up to the previous pixel is 7 pixels, the flat / noise area width data prediction value dst = 7 Is input, the calculated slope data value Ksl is 1/7. When the amplitude limit gradation change amount Xlm is −1, the inclination data value Ksl = −1 / 7. When the flat / noise region width data predicted value dst = 0, the slope data value Ksl = Xlm. Therefore, the slope data value Ksl is obtained as a value in the range from −1 to +1.

  On the other hand, the inclination delay means 40c, when the pixel data of the gray level flat area (Ar = 0) and the noise area (Ar = 1) are input, the non-noise gray level change position one before the current pixel position. The inclination data value Ksl in the pixel of (Ar = 2) is held and output. The slope delay means 40c outputs the slope data value Ksl of the immediately preceding pixel as it is for the pixels in the gray level flat area (Ar = 0) and the noise area (Ar = 1). The value of the slope data value Ksl calculated for each Ar = 2) is held until the next non-noise gradation change position (Ar = 2).

  Therefore, the gradient data value Ksl calculated at the non-noise gradation change position (Ar = 2) is set as the gradient of the gradation value used in the subsequent flat / noise region pixels (Ar = 0, 1). The pixels in the noise region (Ar = 1) as well as the pixels in the flat gradation region (Ar = 0) are treated as pixels in the flat period when calculating the inclination data value Ksl. Therefore, the gradient of the gradation value in the pixel at the non-noise gradation change position (Ar = 2) can be obtained without erroneous detection due to the noise component.

12A and 12B are diagrams for explaining the inclination data value Ksl indicating the inclination of the gradation value obtained by the inclination detecting means 40, and FIG. 12A shows the amplitude limit gradation change amount Xlm. When the gradation value of the gradation increases at +1, FIG. 12B shows a case where the amplitude limit gradation change amount Xlm is −1 and the gradation value decreases. 12A and 12B, the horizontal axis indicates the pixel position in the horizontal direction, and the vertical axis indicates the gradation value La (x) of the input image signal Da at each pixel position. Figure 12 As shown in (a) and (b), non-noise gradation change position (Ar = 2) obtained at the pixel position x ₀ to a flat noise region width data prediction value dst and amplitude limit gradation change From the amount Xlm, a non-noise gradation change position (Ar = 2) indicated by an area signal Ar from the pixel area determination means 20 is set as an inclination as shown by a broken line in FIGS. 12 (a) and 12 (b). pixel position serving _{x 0} since the slope data value Ksl is obtained.
Ksl = Xlm / dst (4)

  As described above, the inclination detecting means 40 outputs the inclination of the change in the gradation value at the non-noise gradation change position (Ar = 2) as the inclination data value Ksl, and becomes a flat / noise area pixel (Ar = 0). , 1) holds the slope value calculated at the previous non-noise gradation change position (Ar = 2). Since the amplitude limit gradation change amount Xlm is limited within the range from −1 to +1, the detected inclination data value Ksl also has a value in the range from −1 to +1.

<< 1-2-7 >> Correction bit generation means 50.
The correction bit generation means 50 includes an inclination data value Ksl from the inclination detection means 40, an area signal Ar from the pixel area determination means 20, and an amplitude restriction from the gradation change amount restriction means 12 of the gradation change amount detection means 10. The gradation change amount Xlm and the correction limiting coefficient rc from the edge change detection means 60 are input. The correction bit generation means 50 obtains the gradation change amount dK for each adjacent pixel of the correction value (correction bit data) to be added to the gradation value of the input pixel signal Da from the inclination data value Ksl, and every horizontal scanning period. At the same time as initialization (set to 0), a correction value aCd is obtained from the gradation change amount dK, the amplitude limit gradation change amount Xlm, and the correction limit coefficient rc in accordance with the area signal Ar, and correction bit data based on the correction value aCd Cd is generated and output to the pixel value calculation means 92. Here, the gradation change amount dK obtained from the inclination data value Ksl is a decimal with respect to the gradation value La (x) (for example, shown in FIGS. 2A and 3A) of the input image signal Da. Therefore, the correction bit data Cd generated by the correction bit generation means 50 includes the additional bit E based on the value including the decimal part of the gradation value of the input image signal Da and the sign bit of the value to be added. Is obtained as a bit string including.

  FIG. 13 is a block diagram schematically showing an example of the configuration of the correction bit generation means 50. As shown in FIG. 13, the correction bit generation unit 50 includes, for example, a calculation unit 51, a subtraction unit 52, a switching unit 53, a correction value delay unit 54, and a bit string conversion unit 55. The correction bit generation means 50 obtains a correction value (gradation value) to be added to the input pixel signal Da, and outputs a bit string indicating the gradation value as correction bit data Cd.

  In FIG. 13, the amplitude limiting gradation change amount Xlm from the gradation change amount limiting means 12 and the correction limit coefficient rc from the edge change detecting means 60 are input to the calculation means 51. The computing means 51 calculates a correction value mCd from the amplitude limit gradation change amount Xlm and the correction limit coefficient rc. Of the correction value mCd obtained by the computing means 51, the correction value mCd in a pixel (a pixel that is not a noise region and whose gradation change amount is not 0) that is a non-noise gradation change position (region signal Ar = 2) is described later. Used by the switching means 53. The correction limiting coefficient rc set by the edge change detecting means 60 is a multiplication value for the gradation correction value, and is set within a range from 0 to 1. Therefore, the correction value mCd is the amplitude limited gradation change amount Xlm. Multiplication is performed by the correction limiting coefficient rc, and the multiplication result (rc) × (Xlm) is calculated.

  The amplitude limit gradation change amount Xlm is limited within a range from −1 to +1 corresponding to the value of one gradation of the input image signal Da, and the correction limit coefficient rc is the magnitude of the gradation change amount df. Therefore, the correction value mCd based on the multiplication result (rc) × (Xlm) is a pixel having a large gradation change amount df and having an edge of the image in the vicinity (correction limiting coefficient). In rc = 0), the correction value mCd = 0, and in other pixels, the absolute value in the range from −1 to +1 is expressed as a decimal that is 1 or less. Further, since it is obtained from the amplitude limit gradation change amount Xlm, the correction value mCd is a value obtained from the height in the slope.

  The correction limit coefficient rc from the edge change detection unit 60 is set as a multiplication value for the correction value, and the calculation unit 51 exemplifies the case where the amplitude limit gradation change amount Xlm is multiplied by the correction limit coefficient rc. By using a ROM or the like that holds a conversion table, the correction value mCd may be obtained from the amplitude limit gradation change amount Xlm according to the value of the correction limit coefficient rc, or according to the value of the correction limit coefficient rc. The correction value mCd may be converted to a fixed value as long as the correction value can be limited according to the correction limiting coefficient rc.

  The subtraction means 52 receives the inclination data value Ksl from the inclination detection means 40 and the delay correction value pCd delayed by one pixel from the correction value delay means 54. The subtracting means 52 obtains the gradation change amount dK for each adjacent pixel of the correction value from the inclination data value Ksl, and subtracts the gradation change amount dK from the correction value in the immediately preceding pixel that is the delay correction value pCd. A correction value sCd is obtained for a pixel that is a flat / noise area after the pixel whose gradation value changes at the non-noise gradation change position (area signal Ar = 2), and the obtained correction value sCd is switched by the switching means 53. Output to.

  That is, the subtracting unit 52 first obtains the value of the gradient data value Ksl as the gradation change amount dK for each adjacent pixel of the correction value (dK = Ksl), and then, the correction value (delay in the immediately preceding pixel). The gradation change amount dK is subtracted (pCd-dK) from the correction value pCd). The subtraction by the subtracting means 52 is subtracted or added according to the sign of the gradation change amount dK (that is, the sign of the slope data value Ksl) (dK> 0 subtracts the absolute value | dK |, dK <0 | DK | is added), and when the sign of the value after the subtraction is inverted from that of the delay correction value pCd (+ changes to-or-changes to +), the subtraction process with the delay correction value pCd being zero In this case, the correction value sCd is output as zero (sCd = 0). As a result, the correction value sCd in the pixel that becomes the flat / noise area after the non-noise gradation change position (Ar = 2) is the magnitude of the absolute value of the gradation change amount dK in order from the change position (that is, the slope data value). The absolute value of Ksl | Ksl |) is changed. Here, since the slope data value Ksl is a value in the range from −1 to +1, the correction value sCd is also expressed as a decimal number whose absolute value in the range from −1 to +1 is 1 or less.

  The area signal Ar, the correction value mCd from the calculation means 51, and the correction value sCd from the subtraction means 52 are input to the switching means 53 in the correction bit generation means 50. The switching unit 53 initializes the value for each horizontal scanning period (sets the fixed correction value to 0), and based on the area signal Ar, the correction value mCd from the calculation unit 51 and the correction value sCd from the subtraction unit 52 To generate a correction value aCd as a gradation value to be added to the input image signal Da and output it.

  More specifically, the switching means 53 selects the initialization value 0 and outputs it as the correction value aCd (= 0) when initialization is indicated at the horizontal scanning period interval, and the non-noise gradation change At the position (Ar = 2), the correction value mCd from the computing means 51 obtained from the amplitude limit gradation change amount Xlm and the correction limit coefficient rc is selected and output as the correction value aCd (= mCd). Further, the switching means 53, in a pixel (Ar = 0, 1) that is a flat gradation area or a noise area, the correction value sCd by the subtraction means 52 (that is, the gradation change amount dK = from the correction value pCd of the immediately preceding pixel). A value obtained by subtracting Ksl) is selected and output as a correction value aCd. The correction value aCd is sent to both the bit string conversion unit 55 and the correction value delay unit 54.

  The correction value delay means 54 delays the correction value aCd from the switching means 53 by one pixel and outputs a delay correction value pCd. Since the correction value delay means 54 delays the output from the switching means 53, the delay correction value pCd becomes the correction value for the immediately preceding pixel.

Due to the configuration of the calculation means 51, subtraction means 52, correction value delay means 54, and switching means 53 described above, the correction value aCd to be added to the input image signal Da is the non-noise gradation change position (Ar = 2). , A correction value mCd based on the amplitude limit gradation change amount Xlm and the correction limit coefficient rc,
mCd = (rc) × (Xlm)
In the pixel (Ar = 0, 1) which is obtained from the above and becomes a gradation flat region or a noise region, the correction value sCd by the subtracting unit 52, that is, the non-noise gradation change position (Ar = 2) in order, The gradation change amount dK = Ksl is subtracted from the value at to obtain a value that sequentially changes to 0 with the level difference of the gradation change amount dK. Both the correction value based on the multiplication result (rc) × (Xlm) and the gradation change amount dK based on the inclination data value Ksl are expressed as decimal numbers whose absolute value in the range from −1 to +1 is 1 or less. The correction value aCd obtained from the switching means 53 is also a gradation value having an accuracy that is a decimal number of 1 or less. Further, since the amplitude limit gradation change amount Xlm corresponds to the height of the gradation change used when calculating the inclination data value Ksl, the correction value aCd is obtained from the inclination value of the gradation value.

That is, in the above, for example, assuming that the value obtained by sequentially counting the pixels of the area signal Ar = 0, 1 after that is the non-noise gradation change position (Ar = 2), the non-noise level The gradation change amount of the correction value aCd in each pixel from the count value N = 0 (correction value mCd) at the tone change position (Ar = 2) is the count value N and the gradient data value Ksl (gradation change amount dK = Ksl) can be expressed by N × Ksl, and the correction value can be expressed by the following equation (5).
aCd = mCd−N × Ksl (5)
Thus, the correction value aCd is a value having a change in the level difference of the gradation change amount dK due to the inclination data value Ksl in order from the position of the correction value mCd.

  By switching the correction value aCd selected by the switching means 53 based on the area signal Ar, the pixel and the noise area (Ar = 1) in the gradation flat area (Ar = 0) after the non-noise gradation change position (Ar = 2). ), The count value N can be obtained to obtain a flat / noise area width, and a correction value aCd having a change in level difference of the gradation change amount dK due to the inclination data value Ksl can be obtained. The correction value aCd that changes sequentially can be obtained without being affected by the above.

  Further, since the correction value mCd in the region signal Ar = 2 is obtained from (rc) × (Xlm), the pixel value of the pixel whose gradation value changes gradually and gradually in the horizontal direction between +1 and −1. In the region, the correction value aCd has a change in the level difference of the gradation change amount dK in order from the change position of the gradation value. On the other hand, when it is determined that the change in the gradation value is due to an edge, When the correction limiting coefficient rc = 0, the correction value mCd = 0, and the correction value is 0 even in the subsequent pixels. Therefore, no gradation value is added. Therefore, it is possible to obtain a correction value that maintains the sharpness (contour information) of the image.

  The correction value aCd from the switching unit 53 is input to the bit string conversion unit 55 in the correction bit generation unit 50. The bit string converting means 55 converts the correction value aCd into a bit string value having a predetermined number of bits, obtains the additional bit E indicating the value to be added and the sign bit of the value as corrected bit data Cd, and generates a corrected bit generating means 50 to the pixel value calculation means 92. Since the absolute value excluding the sign of the correction value aCd is obtained as a value including a decimal number of 1 or less, the correction value (additional bit E) in the correction bit data Cd is the bit string of the gradation value of the input image signal Da. Indicates the decimal part.

  Here, when the additional bit E in the correction bit data Cd is generated with 4 bits and added to the lowest order side of the input image signal Da, for example, the absolute value of the correction value aCd excluding the sign is the value of a 4-bit bit string. Convert to However, at this time, the value that becomes the integer part 1 of −1 and +1 is 16 in the bit string. Therefore, this value is limited to 4 bits, the value in the bit string is obtained as 15, and the bit string is obtained. The conversion means 55 converts the correction value aCd into a bit string value having a predetermined number of bits, and also performs a value limiting process. Then, together with a sign bit (having a bit number of 1 bit or more) indicating the sign of the correction value aCd, the converted 4-bit additional bit E is generated as correction bit data Cd and output.

  The correction bit generation means 50 of FIG. 13 has been described to provide the bit string conversion means 55 to obtain the correction value aCd as correction bit data Cd by the bit string of the sign bit and the 4-bit additional bit E. Since the value such as the correction value aCd is usually obtained as a value of a predetermined number of bits, it can also be used as the correction bit data Cd as it is. The bit string conversion means 55 can be omitted.

<< 1-2-8 >> Delay compensation means 91.
When the input image signal Da is input to the delay compensation unit 91, the delay compensation unit 91 matches the delay until the correction bit data Cd output from the correction bit generation unit 50 is obtained and the pixel position of the input image signal Da. Thus, the delay compensation of the input image signal Da is performed, and the delay-compensated image signal Dd is output.

<< 1-2-9 >> Pixel value calculation means 92.
The pixel value calculation unit 92 receives the delay compensated image signal Dd from the delay compensation unit 91 and the correction bit data Cd from the correction bit generation unit 50. The pixel value calculation unit 92 calculates a new output image signal Db by subtracting the correction bit data Cd from the correction bit generation unit 50 by matching the position of the integer part with respect to the delay-compensated image signal Dd. ,Output. Since correction bit data Cd including a decimal part is added to the gradation value of the output image signal Db by calculation, the number of bits of the output image signal Db is larger than the number of bits of the image signal Dd, that is, expanded. The output image signal Db in which the number of bits is expanded becomes the output of the gradation converting means 1.

  The pixel value calculation unit 92 performs subtraction or addition processing according to the sign of the correction bit data Cd in the calculation processing for subtracting the correction bit data Cd from the image signal Dd (that is, when Cd> 0, The absolute value | Cd | is subtracted, and if Cd <0, | Cd | is added). Further, the pixel value calculation unit 92 sets the output image signal Db to “0” (Db = 0) when the calculated value is a negative value, and sets the number of bits (maximum value) set in the output image signal Db. If the value exceeds the value, the value is set to the maximum value, and the value of the output image signal Db is limited.

  As described above, the additional bit E indicating the value to be added in the correction bit data Cd includes a fractional part with respect to the gradation value of the delay-compensated image signal Dd (or input image signal Da). The output image signal Db after being subjected to arithmetic processing by addition or subtraction in the arithmetic means 92 becomes an image signal with a precision having a decimal part with respect to the image signal Dd. That is, since the number of bits of the output image signal Db is larger than that of the image signal Dd, not only the number of gradations that can be expressed is increased, but also the number of bits in the decimal part by the correction bit Cd including the decimal part is included. The resolution of the tone value is increased, and a smoother gradation value change can be expressed. When the change in the gradation value is caused by the edge of the image, the correction bit data Cd is obtained as 0. As a result, no gradation is added to the pixel adjacent to the edge portion of the image. While maintaining the sharpness (contour information) of the signal Db, it is possible to increase the smoothness of the image in the pixel region where the gradation value gradually changes other than the edge of the image.

  Further, the correction bit data Cd is based on the region signal Ar from the pixel region determination unit 20 and is not affected by the noise component, and the gradation data value Ksl is sequentially generated from the non-noise gradation change position (Ar = 2) as a starting point. It is obtained as a correction value aCd that changes with the tone change amount dK. In the pixel value calculation unit 92, the value of the correction bit data Cd is added to the gradation value of the image signal Dd (or the input image signal Da). Subtracted. For this reason, compared to tone value conversion processing by linear interpolation processing, the pixel area where the tone value changes gradually while keeping the image detail and sharpness due to slight changes in tone value such as noise remains smooth. Thus, the image blur can be reduced.

  FIG. 14 shows the input image signal Da (or the delay compensated image signal Dd from the delay compensation unit 91), the correction bit data Cd from the correction bit generation unit 50, and the output image output from the pixel value calculation unit 92. It is a figure which shows the relationship of the signal Db. Here, a case will be described where the input image signal Da (or image signal Dd) is m = 8 bits, the correction bit data Cd is a sign bit and an additional bit E4 bits, and the output image signal Db is n = 12 bits.

  In FIG. 14, since the correction bit data Cd is obtained as an additional bit E indicating a decimal part with respect to the input image signal Da, the pixel value calculation means 92 aligns the position of the integer part and the lowest order of the image signal Dd. The output image signal Db is calculated by subtracting or adding a value obtained by adding the additional bit E (4 bits) of the correction bit data Cd to a position lower than the bit according to the sign. That is, the pixel value calculation unit 92 performs a process corresponding to adding the value of the additional bit E to the least significant side of the image signal Dd, and generates a 12-bit output image signal Db. The output image signal Db has a bit string extended to the lower bit side with respect to the image signal Dd, that is, includes a fractional part with respect to the gradation value of the original input image signal Da. The resolution of gradation is higher than that of the image signal Dd (16 times in FIG. 14), quantization noise or pseudo contour is further reduced, and an accurate gradation value can be expressed.

<< 1-3 >> Operation.
Next, in the image processing apparatus according to the first embodiment, the gradation change amount df is obtained from the difference in gradation value between the pixels of the input image signal Da, and each pixel is flattened based on the gradation change amount. Performs determination to classify the pixel into one of a pixel belonging to a region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position, and the flatness / noise region width and inclination in gradation change from the region determination result and the gradation change amount df The operation of generating the corrected bit data, and adding the corrected bit data to the input image signal Da to output the output image signal Db in which the number of gradations is expanded will be described with reference to the flowchart.

  FIG. 15 is a flowchart showing the operation (part 1) of the image processing apparatus according to the first embodiment, and FIG. 16 is a flowchart showing the operation (part 2) of the image processing apparatus according to the first embodiment. In the image processing apparatus according to the first embodiment, the correction bit data Cd is added to the input image signal Da to expand the number of bits and perform gradation conversion to obtain the gradation value of the original input image signal Da. On the other hand, an output image signal Db with a decimal part added is output. FIG. 17 is a flowchart showing the operation of the pixel area determination means 20 shown in FIG. In the processing in the gradation converting means 1, initialization is performed for each horizontal synchronization signal. However, in FIGS. 15 to 17, the description of the operation initialized for each horizontal synchronization signal is omitted, and each scan is performed. This is shown as processing for each pixel in the line.

  FIGS. 18A to 18K show the operation of the gradation converting means 1 for a part of the input image signal Da including the noise component shown in FIG. 3A, and FIG. The pixel position x in the horizontal direction of the input image signal Da, (b) in the figure is the input image signal Da, and (c) and (d) in FIG. The gradation change amount df due to the difference between the values and the amplitude-limited gradation change amount Xlm of the gradation change amount df, FIG. 9E shows the region signal Ar obtained by the pixel region determination means 20, and FIG. Flatness in the noise region width determination means 30 Flatness that is an output from the noise region width count means 31. Noise area width count value Cnt, FIG. The flat / noise area width output from the processing means 32 The measured flat / noise area width data predicted value dst, FIG. 11H shows the slope data value Ksl output from the slope detection means 40, and FIG. 11I shows the correction value aCd obtained by the correction bit generation means 50. FIG. 6 (j) shows the gradation when the value obtained by adding the correction bit data Cd to the image signal Dd (that is, the input image signal Da) in the pixel value calculation means 92 is represented by 1 as the minimum gradation value of the image signal Dd. (K) in the figure shows the output image signal Db which is the output of the pixel value calculation means 92, and shows the case where the additional bit E in the correction bit data Cd is 4 bits. In FIGS. 18A to 18K, pixel data is input in order from the left, and the horizontal axis indicates the pixel position in the horizontal scanning direction. 18A to 18K, in order from the pixel position x = 8, there are 8 pixels with gradation value La = 2, 8 pixels with gradation value La = 3, and gradation value La = 4. When the input image signal Da is m = 8 bits, 8 gradations are arranged side by side with respect to the 8-bit input image signal Da, one gradation from the left pixel position x = 8 to the right. It is an image that becomes brighter continuously, and includes high-frequency changes that become noise components at pixel positions x = 19 and x = 34, 35 during the gradation change.

  FIGS. 19A to 19K show examples of signals that include noise components and in which the level of the input image signal Da gradually decreases by one gradation, and FIGS. The signals in k) are the same as the signals in FIGS. 18 (a) to 18 (k), respectively. In FIGS. 19A to 19K, the gradation value La = 1 to 8 changes at the pixel position x = 8, and the gradation value La = 8 is sequentially changed at the subsequent pixel position x. There are 8 pixels, 8 pixels with a gradation value La = 7, 8 pixels with a gradation value La = 6, and so on. The pixel position x = 8 with respect to the 8-bit input image signal Da. The change of the gradation difference 7 at 8 is an image that becomes darker continuously by one gradation after that due to the contour (edge) of the image, and the noise component at the pixel positions x = 19 and x = 35 during the gradation change. It includes a high frequency change.

  In the following, the operation of the image processing apparatus (also the operation of the gradation converting means 1 in the first embodiment) is shown in FIGS. 15 to 17, FIGS. 18 (a) to (k), and FIG. 19 (a). It demonstrates using-(k). First, with reference to FIG. 15, a description will be given of obtaining the gradation change amount df and obtaining the flat / noise area width and inclination in the gradation change.

When an m-bit input image signal Da (referred to as a gradation value La (x)) as shown in FIGS. 18B and 19B is input to the gradation converting means 1 (step S101), This input image signal Da is input to the gradation change amount detection means 10. Then, the gradation change amount calculation means 11 in the gradation change amount detection means 10 determines the difference between the gradation values of pixels adjacent in the horizontal direction (that is, the gradation value of the pixel at the immediately preceding pixel position x−1). Difference), and the gradation change amount df = La (x) −La (x−1)
Is obtained (step S102). For example, in FIG. 18C, the gradation change amount df = La (8) −La (7) at the pixel position x = 8 is calculated, and df = + 1. Further, for example, in FIG. 19C, the gradation change amount df = La (8) −La (7) at the pixel position x = 8 is calculated to be df = + 7, and at the pixel position x = 16, the level is changed. The tone change amount df = −1.

  The gradation change amount df obtained by the gradation change amount calculation means 11 is limited to a predetermined range by the gradation change amount restriction means 12 in the gradation change amount detection means 10, and the amplitude restriction gradation A change amount Xlm is obtained (step S103). The amplitude limit gradation change amount Xlm is limited so that the gradation change amount df is within a predetermined level, that is, within a range from −1 to +1 (indicated by −TH to + TH in FIG. 15). This is a limit range for detecting a change in one gradation of the input image signal Da. That is, the value of the amplitude limit gradation change amount Xlm is represented by three values of −1, 0, and +1. The amplitude-limited gradation change amount Xlm is used as the height of the gradation change when obtaining the gradient of the gradation value. The amplitude-limited gradation change amount Xlm in FIGS. 18D and 19D is +1 or −1 in the pixel where there is a change due to a position where the gradation value changes or a noise component. For a pixel in a flat gradation region that is limited in value and has no change, the amplitude limited gradation change amount Xlm = 0.

  The gradation change amount df is also output to the edge change detection means 60. The edge change detection means 60 detects a change that becomes an edge of the image from the magnitude of the absolute value of the gradation change amount df, and sets a correction limiting coefficient rc (step S104). Here, the correction limitation coefficient rc to be obtained is obtained as a value that can restrict the correction value so that the gradation value is not added at the edge portion. It is set with a value in the range up to 1.

  Since the gradation change amount df is obtained from the difference in gradation value between adjacent pixels, the pixel near the edge of the image includes the difference in gradation value due to the edge component, but the pixel that expands the number of gradations In this area, the gradation value changes smoothly and is an area that changes with a difference of one gradation. Therefore, it is necessary to prevent the correction value from being added to the edge portion. Therefore, it is determined from the magnitude of the gradation change amount df whether or not the difference in gradation value between pixels adjacent to the edge is due to the edge, and the correction limiting coefficient rc is set based on the result. As an example, when the correction limiting coefficient rc is determined as rc = 0 to 1 by converting from the magnitude of the absolute value of the gradation change amount df with the characteristics shown in FIG. 9, the gradation change amount df becomes an edge. In the pixel detected as a change due to the above, the value of the correction limiting coefficient rc is rc = 0, and the correction value is limited by the value.

  In FIGS. 18A to 18K, the gradation change amount df (FIG. 18A) has an absolute value | df | ≦ 1 at any position, so that the correction limiting coefficient rc is always the correction limiting coefficient rc. = 1 (not shown). In the case of FIGS. 19A to 19K, the gradation change amount df (FIG. 19A) at the pixel position x = 8 is df = + 7, and the correction limiting coefficient rc at this position is rc = 0. However, at other pixel positions, the correction limiting coefficient rc is rc = 1.

  Next, the pixel region determination unit 20 determines each gradation value from the change in the gradation value in the target pixel at the pixel position x by performing determination using the value and polarity of the gradation change amount df from the gradation change amount calculation unit 11. It is determined whether the pixel is classified into any of the three areas of the flat gradation area, the noise area, and the non-noise gradation change position, and an area signal Ar indicating the pixel area determination result is generated (step S105). . The detailed operation of determining the pixel area based on the gradation change amount df will be described later with reference to FIG. 17, but the noise component is a high-frequency change that increases and decreases in a range of several pixels. Since the tone difference is not more than a relatively small value Nd (for example, Nd = 2), the change in the pixel due to the noise component is the value of the gradation change amount df in the target pixel at the pixel position x and the value immediately before the target pixel. It can be determined by the polarity of the gradation change amount df in each of the adjacent pixel at the pixel position x-1 and the adjacent pixel at the pixel position x + 1 immediately after the target pixel. A determination is made to classify the pixel into either a pixel to which it belongs or a pixel at a non-noise gradation change position (FIGS. 18E and 19E).

  The flat / noise area width counting means 31 in the flat / noise area width determining means 30 is provided for each pixel in the horizontal scanning period up to the pixel immediately before the non-noise gradation change position (Ar = 2) with reference to the area signal Ar. The number of pixels of the flat / noise area width is obtained and set as the flat / noise area width count value Cnt (step S106). That is, initialization is performed for each horizontal scanning period, and the widths of pixels that become gradation flat regions (Ar = 0) and noise regions (Ar = 1) between the non-noise gradation change positions (Ar = 2) ( The number of pixels) is counted, and the flat / noise area width up to the pixel immediately before the non-noise gradation change position (Ar = 2) is obtained as the flat / noise area width count value Cnt.

  In FIG. 18F, the pixel count is reset to 0 at the pixel position x = 8, 16, 24, 32 which is the non-noise gradation change position (Ar = 2), and at the next pixel position (x + 1). The flat / noise area width count value Cnt becomes zero. At the pixel position x that is the non-noise gradation change position (Ar = 2), the flat / noise area width that is the number of pixels counted up to the previous pixel position x−1, that is, the gradation flat area (Ar = 0). ) And a flat / noise area width count value Cnt indicating the continuous number of pixels to be a noise area (Ar = 1). The same applies to FIG.

  Next, in the flat / noise area width determining means 30 and the inclination detecting means 40, an operation for obtaining a flat / noise area width data predicted value dst, which is a predicted value of the flat / noise area width, and detecting the gradient of the gradation value. Is performed at the non-noise gradation change position of the area signal Ar = 2 (steps S107 to S110).

At the non-noise gradation change position where the area signal Ar = 2 (step S107), the recursive filter processing means 32 determines the non-noise gradation change position (Ar = 2) with respect to the flat / noise area width count value Cnt. A cyclic filter process is performed every time, and the result of the filter process is a flat / noise area width data predicted value dst that is a value obtained by predicting the flat / noise area width after the non-noise gradation change position (Ar = 2). (Step S108). The calculation of the recursive filter processing constitutes a low-pass filter, and for the delay flat / noise region width data predicted value pdst and the flat / noise region width count value Cnt that are filter outputs, in FIG. Cyclic filter processing is performed using (pdst + Cnt) / 2 (coefficient k ₁ = 1/2, k ₂ = 1/2 in the above equation (1)). On the other hand, in the gradation flat region (Ar = 0) and noise region (Ar = 1), the value at the previous non-noise gradation change position (Ar = 2) is held and output (step S109).

  The flat / noise area width data predicted value dst is a low frequency component of the flat / noise area width when the gradation value changes in the horizontal direction. Therefore, in the pixel region where the gradation value changes continuously and gradually in the horizontal direction, the width of the gradation flat region when the gradation value changes can be obtained while avoiding a steep change. Thus, the flat / noise area width after the non-noise gradation change position (Ar = 2) can be used as a predicted value. Further, since the flat / noise area width data predicted value dst is obtained from the flat / noise area width count value Cnt obtained by counting the pixels with reference to the non-noise gradation change position (Ar = 2), the floor where the area signal Ar = 0 is obtained. The pixels in the noise region indicated by the region signal Ar = 1 as well as the pixels in the tonal flat region are treated as pixels in the flat period, and the noise region is included (that is, without erroneous detection due to the noise component) can get.

  18 (g) and 19 (g), flat / noise area width data prediction at x-1 one pixel before at the non-noise gradation change position (Ar = 2) x = 8, 16, 24, 32. A cyclic filter process of (pdst + Cnt) / 2 is performed on the value pdst and the flat / noise area width count value Cnt (FIG. 18 (f) and FIG. 19 (f)). In the flat gradation region (Ar = 0) and the noise region (Ar = 1), values at pixel positions x = 8, 16, 24, and 32 are held.

  In accordance with the area signal Ar, the inclination detecting means 40 in the horizontal direction of the input image signal Da from the amplitude limited gradation change amount Xlm and the flat / noise area width data predicted value dst at the non-noise gradation change position (Ar = 2). Is detected as a gradient data value Ksl = Xlm / dst (steps S107 and S110). In the gradation flat region (Ar = 0) and the noise region (Ar = 1), the value at the previous non-noise gradation change position (Ar = 2) is held (step S109). Since the amplitude limit gradation change amount Xlm is input as a value in which the gradation change amount df is limited within a range from −1 to +1 (step S103), the gradient of the detected gradation value, that is, the gradient data value. Ksl is also a value within a range from −1 to +1.

  According to FIG. 18H, at the non-noise gradation change position (Ar = 2) x = 8, the inclination data value Ksl = + 1/4, and similarly, the inclination data at the pixel position x = 16, 24, 32. A value Ksl is determined. Further, according to FIG. 19H, the slope data value Ksl = + 1/4 at the non-noise gradation change position (Ar = 2) x = 8, and the slope data value Ksl at the pixel position x = 16. = -1 / 6, and similarly, the inclination data value Ksl is obtained at the pixel position x = 24,32. In both FIG. 18 (h) and FIG. 19 (h), the values at the pixel positions x = 8, 16, 24, and 32 are held in the flat gradation region and the noise region (Ar = 0, 1). Yes.

  With the operation shown in FIG. 15, the flat / noise area width data predicted value dst, which is a value obtained by predicting the flat / noise area width in the gradient of the gradation value with reference to the area signal Ar, and the gradient of the gradation value are obtained. The gradient of the gradation value in the pixel at the non-noise gradation change position (Ar = 2) can be obtained without erroneous detection due to the component. Further, the edge change detection means 60 determines whether or not the difference in gradation change is due to the edge from the magnitude of the gradation change amount df, and generates a correction limiting coefficient rc from the result.

  Here, an operation of determining the pixel area based on the gradation change amount df in the pixel area determining unit 20 will be described with reference to FIG.

  The pixel area determination unit 20 first delays the pixel with respect to the gradation change amount df from the gradation change amount detection unit 10, so that the pixel of interest at the pixel position x and the adjacent pixel at the pixel position x−1 immediately before it are detected. The gradation change amount df in each of the adjacent pixels at the pixel position x + 1 immediately after the target pixel is obtained as the gradation change amounts df (x), df (x−1), and df (x + 1) (step S120).

  Next, the polarities of the gradation change amount df (x) and the gradation change amounts df (x), df (x−1), and df (x + 1) are acquired (step S121), and the pixel position x is obtained. The target pixel has a high-frequency component change in an adjacent pixel in the noise region (the polarity of the tone change amount df (x) is opposite to the polarity of the tone change amount df (x−1), or the tone change It is determined whether the polarity of the amount df (x) is opposite to the polarity of the gradation change amount df (x + 1) (steps S122 to S129). Specifically, the adjacent pixel which is one condition for determining the noise region has a high frequency component, that is, the polarity of the gradation change amount df (x) is the gradation change amount df (x−1). Or the polarity of the gradation change amount df (x + 1).

  First, the polarity of the gradation change amount df (x) and the gradation change amount df (x−1) are determined as to whether there is a change in the high frequency component whose direction of change is opposite at the pixel position x−1 adjacent to the pixel position x. ) From the polarity (step S122). When the polarity of the gradation change amount df (x) is opposite to the polarity of the gradation change amount df (x−1) (step S124), the determination result pf1 = 1 indicating a binarized value is set (step S124). In S126), if they are the same, the determination result pf1 = 0 is set (step S127).

  Similarly, the polarity of the gradation change amount df (x) and the polarity of the gradation change amount df (x + 1) indicate whether there is a change in the high-frequency component whose direction of change is opposite between the pixel position x and the adjacent pixel position x + 1. (Step S123). When the polarity of the gradation change amount df (x) and the polarity of the gradation change amount df (x + 1) are opposite (step S125), the determination result pf1 = 2 indicating the binarized value is set (step S128). If they are the same, the determination result pf2 = 0 is set (step S129).

  Then, the level of the absolute value of the gradation change amount df (x) in the target pixel at the pixel position x is compared with a predetermined threshold value. Is determined to be less than or equal to a relatively small value Nd, and by using the determination result of the polarity, each pixel is classified into a pixel belonging to a flat gradation region, a pixel belonging to a noise region, and a non-noise gradation by the following operation. A determination is made to classify the pixel into one of the change position pixels, and an area signal Ar indicating each area determination result is generated.

(Decision A1)
When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0 (step S130), the target pixel is a pixel belonging to the flat gradation region, and the region signal Ar Becomes Ar = 0 (step S133).

(Decision A2)
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd (steps S130 and S131), and the polarity of the gradation change amount df (x) is When the polarity is opposite to the polarity of the gradation change amount df (x−1) (when the adjacent pixel has a high frequency component) (pf1 = 1 in step S132), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd (steps S130 and S131), and the polarity of the gradation change amount df (x) is When the polarity is opposite to the polarity of the gradation change amount df (x + 1) (when the adjacent pixel has a high frequency component) (pf2 = 1 in step S132),
The target pixel is a pixel belonging to the noise region, and the region signal Ar becomes Ar = 1 (step S135).

(Decision A3)
The pixel of interest x at the pixel position x determined that the target pixel is not a pixel belonging to the flat gradation region in the determination A1 and is determined not to be a pixel belonging to the noise region in the determination A2. The pixel at the noise gradation change position, and the area signal Ar becomes Ar = 2 (step S134).

  The predetermined threshold Nd is set to a value that is sufficiently smaller than the image signal and takes into account the magnitude of the amplitude of noise components such as noise and error contained in the input image signal. Nd = 2.

  In FIG. 18 (e), the pixel position x = 8, 16, 24, 32 is determined as the area signal Ar = 2 at the non-noise gradation change position because the gradation change amount df = 0 of the adjacent pixel. The At the pixel position x = 19 where the gradation change amount df is not 0 due to the noise component and the adjacent pixel position x = 20, the polarity of the gradation change amount df is opposite to each other (df (19) = + 1, df (20 ) = − 1) Therefore, it is determined as a noise region, and the region signal Ar = 1. Similarly, x = 34, 35, and 36 are also determined as noise regions. The same applies to FIG. 19F, and the pixel positions x = 8, 16, 24, and 32 are not flat gradation regions because of df ≠ 0, and the gradation change amount df = 0 of the adjacent pixels. Since it is not a noise region, it is a non-noise gradation change position, and the region signal Ar = 2. The pixel at pixel position x = 19 and its adjacent pixel (pixel position x = 20) and the pixel position x = 35 and its adjacent pixel (pixel position x = 36) where the gradation change amount df ≠ 0 due to the noise component ), Since the polarities of the gradation change amounts df are opposite to each other, it is determined as a noise region, and the region signal Ar = 1.

  As described above, the operation shown in FIG. 17 determines that each pixel is classified as one of a pixel belonging to a flat gradation region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position. , An area signal Ar is obtained, and it is indicated which gradation value of each pixel has a gradation flat area, a noise area, and a non-noise gradation change position.

  Next, the operation of expanding the number of bits by generating corrected bit data and adding it to the input image signal Da will be described with reference to FIG. In FIG. 16, after the gradient data value Ksl, which is the gradient of the gradation value, is obtained in accordance with the area signal Ar, the correction bit generation unit 50 initializes the correction value aCd at the boundary of the horizontal scanning period and The process of generating the correction value aCd is switched according to the area signal Ar indicating the area (step S111).

  At the non-noise gradation change position (Ar = 2) (step S111), the amplitude limit gradation change amount Xlm is multiplied by the correction restriction coefficient rc, and the correction value mCd as the multiplication result is (rc) × (Xlm). (Step S112). Then, the obtained correction value mCd is determined as the correction value aCd (step S113). The amplitude limit gradation change amount Xlm is limited within a range from −1 to +1 corresponding to one gradation value of the input image signal Da, and the correction limit coefficient rc is larger than the gradation change amount df. Since the correction value mCd is obtained with a value of 0 to 1 according to the above, the correction value mCd = 0 for a pixel (correction limiting coefficient rc = 0) in which the gradation change amount df is large and the edge exists in the vicinity. In other pixels, the absolute value in the range from −1 to +1 is expressed as a decimal with 1 or less. Since the amplitude limit gradation change amount Xlm is the height of the gradation change used when calculating the gradient of the gradation value, the correction value obtained is obtained from the gradient of the gradation value.

  On the other hand, in the pixel that becomes the gradation flat region (Ar = 0) and the noise region (Ar = 1) (step S111), the gradient data value Ksl is set as the gradation change amount dK for each adjacent pixel of the correction value (dK). = Ksl), the gradation change amount dK is subtracted (pCd−dK) from the correction value (delay correction value pCd) of the immediately preceding pixel to obtain the correction value sCd of the pixel in the flat gradation region (step S114). After subtraction, when the sign of the value of the correction value sCd is inverted from the delay correction value pCd (+ changes to − or − changes to +), and in the case of subtraction processing when the delay correction value pCd is zero, the correction value sCd is output as zero (sCd = 0). Then, the correction value sCd is determined as the correction value aCd (step S115). As a result, the correction value sCd in the pixels of the flat gradation region (Ar = 0) and the noise region (Ar = 1) is the magnitude of the absolute value of the gradation change amount dK in order from the change position (that is, the slope data value Ksl). The absolute value of | Ksl |) is changed. Here, since the slope data value Ksl is a value in the range from −1 to +1, the correction value sCd is also expressed as a decimal number whose absolute value in the range from −1 to +1 is 1 or less.

  The correction value aCd obtained as described above is converted into a bit string value having a predetermined number of bits, and is determined as correction bit data Cd including an additional bit E indicating a value to be added and a sign bit of the value (step S116). ). The correction value aCd is delayed by one pixel and sent to the next pixel process as the correction value pCd in the immediately preceding pixel (step S117).

  Through the above steps S111 to S117, the correction value aCd (that is, the correction bit data Cd) to be added to the input image signal Da is set to the amplitude limit gradation change amount Xlm at the non-noise gradation change position (Ar = 2). The correction value aCd = (rc) × (Xlm) obtained by the correction limiting coefficient rc (that is, the height of the gradation change used when calculating the inclination is obtained), and the gradation flat region (Ar = 0) In the noise region (Ar = 1), the values are sequentially changed from the value at the non-noise gradation change position (Ar = 2) to a value that sequentially changes to 0 due to the level difference by the slope data value Ksl. Both the correction value based on the multiplication result (rc) × (Xlm) and the slope data value Ksl are expressed as decimal numbers whose absolute value in the range from −1 to +1 is 1 or less, and therefore the correction value aCd is also expressed. 1 is a gradation value having an accuracy of a decimal number of 1 or less.

  Since the correction value aCd to be obtained is switched based on the area signal Ar, the pixel that becomes the noise area (Ar = 1) as well as the gradation flat area has a change in level difference due to the inclination data value Ksl as a pixel in the flat period. Correction values aCd are obtained, and correction values aCd that change sequentially are obtained without being affected by noise components.

  In addition, since the correction value mCd is obtained from (rc) × (Xlm) using the correction limiting coefficient rc, in the region of the pixel where the gradation value changes gradually, the gradation is changed in order from the non-noise gradation change position. The correction value aCd has a change in the level difference of the tone change amount dK. On the other hand, if it is determined that the change in the gradation value is due to an edge, the correction value mCd = 0, and the correction value is also applied to the subsequent pixels. Becomes 0, and can be obtained as a correction value for maintaining the contour information of the image.

  According to FIG. 18 (i), the correction value aCd = + 1 at the pixel position x = 16 (Ar = 2), and the pixel position x = 16 in the pixels that become the gradation flat region and noise region thereafter. The correction value aCd having a change of the level difference −1/6 by the inclination data value Ksl (that is, the correction bit data), such as 5/6, 4/6,. Cd). Similarly, at the pixel position x = 32, the correction value aCd = + 1, and in the pixels that become the gradation flat region and the noise region after that, the correction value aCd = + 1 at the pixel position x = 32 in order from 6 / 7, 5/7, 4/7,... Is obtained as a correction value aCd having a change of level difference −1/7 by the inclination data value Ksl.

  Further, according to FIG. 19 (i), the correction value aCd = −1 at the pixel position x = 16 (Ar = 2), and the pixel that becomes the gradation flat region and the noise region after that is the pixel Correction with a change of level difference + 1/6 due to the inclination data value Ksl, such as −5/6, −4/6, −3/6,... Sequentially from the correction value aCd = −1 at the position x = 16. Obtained as a value aCd (that is, corrected bit data Cd). At the pixel position x = 8, the correction value aCd = 0 because the gradation change amount df = + 7 and the correction limiting coefficient rc = 0. Therefore, the correction value aCd (that is, the correction bit data Cd) is 0 even in the subsequent flat pixels.

  The pixel value calculation unit 92 aligns the position of the integer part of the correction bit data Cd with respect to the image signal Dd (or input image signal Da) that has been delay compensated so as to match the pixel position of the correction bit data Cd and the input image signal Da. By subtracting, a new output image signal Db is calculated (step S118). That is, in the calculation processing of the correction bit data Cd for the image signal Dd, the value of the correction bit data Cd is subtracted from the gradation value La of the image signal Dd, and the value of the additional bit E is combined with the least significant side of the image signal Dd. A process corresponding to the addition is performed (see FIG. 14). When the value after the calculation is a negative value, the output image signal Db is set to zero (Db = 0), and when the value exceeds the set bit number (maximum value) of the output image signal Db, Is the maximum value, and the value of the output image signal Db is limited.

  Since the correction bit data Cd includes a decimal part with respect to the gradation value of the image signal Dd (or the input image signal Da), the output image signal Db after the arithmetic processing has a decimal part with respect to the image signal Dd. It becomes an image signal with a certain accuracy. Since the output image signal Db has the bit number of the decimal part by the correction bit data Cd including the decimal part, the resolution of the gradation value is increased, and a smoother change of the gradation value can be expressed. The correction bit data Cd is obtained as a correction value aC that changes with the gradient data value Ksl without being affected by the noise component with reference to the area signal Ar, and the value of the correction bit data Cd is added to the gradation value or Since the subtraction processing is performed, the smoothness of the pixel region in which the gradation value gradually changes can be increased without reducing the sharpness while leaving the image detail due to a slight change in the gradation value such as noise. it can.

  According to FIG. 18J, the gradation value Lb of the output image signal Db after adding the correction bit data Cd is obtained by subtracting the correction bit data Cd from the gradation value La of the image signal Dd. At the gradation change position (Ar = 2) x = 16, the gradation value Lb = 12/6, and in the gradation flat area and the noise area after that, the change increases by 1/6 in order of the pixel position. In addition, at a pixel position x = 19 where there is a noise component, the gradation value is obtained as a converted value 21/6 while leaving the amplitude (6/6) due to the noise component. Similarly, at the pixel position x = 32, the gradation value Lb of the output image signal Db is 28/7, and the subsequent gradation flat region and noise region have a change that increases in order of 1/7. In the pixel position x = 34 where there is a noise component, the gradation value is obtained as a converted value 37/7 while leaving the amplitude (7/7) due to the noise component. The same applies to the pixel position x = 35.

  Further, according to FIG. 19 (j), the gradation value Lb = 48/6 at the non-noise gradation change position (Ar = 2) x = 16, and the gradation flat area and the noise area after that are 1 in order. In the case of x = 19 where there is a change that decreases at every / 6 and there is a noise component, the gradation value is obtained as a converted value 51/6 while leaving the amplitude (6/6) due to the noise component. Similarly, at the pixel position x = 32, the gradation value Lb of the output image signal Db is 42/7, and the subsequent gradation flat region and noise region have a change that decreases in order of 1/7. In the pixel position x = 35 where there is a noise component, the gradation value is obtained as a converted value 32/7 while leaving the amplitude (7/7) due to the noise component. Note that, at the pixel position x = 8 in FIG. 19J, the correction value aCd (that is, the correction bit data Cd) is 0, so the pixel position x = 8 and the subsequent flat pixels (up to x = 15). No change in gradation value is possible even in the case of the pixel.

  FIGS. 18 (k) and 19 (k) show the output image signal Db when the additional bit E in the correction bit data Cd is 4 bits, and the pixel from the non-noise gradation change position (Ar = 2). A 12-bit output image signal Db obtained by converting the gradation change more smoothly is obtained as a gradation value having a change corresponding to the inclination data value Ksl in the order of the position.

  1, 14, 18, and 19 show a case where the additional bit E in the correction bit data Cd is generated with 4 bits, but the additional bit E is not limited to 4 bits, and more By expressing the correction bit data Cd (that is, the correction value aCd) with the number of bits, it is possible to obtain an image signal with a high-precision and smoother gradation value change. For example, in FIGS. 18A to 18K, since the inclination data value Ksl is +1/4 from the pixel position x = 8 to the pixel position where the next gradation value changes, the correction value aCd In order to show the change of the absolute value 1/4, if the additional bit E is composed of 3 bits, the corrected bit data Cd can be expressed without error. However, after the pixel position x = 24 in FIGS. 18A to 18K, since the slope data value Ksl is +1/7, it indicates the gradation change amount in the correction value aCd, and therefore more as additional bits E. Therefore, the corrected bit data Cd can be expressed with a small error.

  In other words, the correction value aCd is obtained from the gradient data value Ksl obtained from the amplitude limit gradation change amount Xlm and the flat / noise region width data predicted value dst, and thus the maximum width in the flat / noise region width data predicted value dst. The change in the correction value aCd can be made finer by increasing (the number of pixels) and increasing the number of bits representing the inclination data value Ksl. Therefore, if the number of additional bits E in the correction bit data Cd is increased, an image with higher resolution (accuracy) of gradation values can be obtained.

<< 1-4 >> Effect.
As described above, in the image processing apparatus or the image processing method of the first embodiment, using the gradation change amount df, each pixel is classified into a pixel belonging to the gradation flat area, a pixel belonging to the noise area, A determination to classify the pixel into one of the noise gradation change positions is performed, and a flat / noise area width after the non-noise gradation change position (Ar = 2) is predicted based on the area signal Ar indicating the determination result. The flat / noise area width data prediction value dst is obtained, and the gradient (gradient) of the gradation value is calculated from the amplitude-limited gradation change amount Xlm at the non-noise gradation change position (Ar = 2) and the flat / noise area width data prediction value dst. The data value Ksl) is obtained. In the image processing apparatus or the image processing method according to the first embodiment, when it is determined that the gradation change is caused by the edge of the image, the correction limiting coefficient rc for limiting the correction in the correction bit generation unit 50 is set. The correction bit data Cd is generated from the amplitude limit gradation change amount Xlm at the non-noise gradation change position (Ar = 2), the inclination data value Ksl, and the correction limit coefficient rc, and the position of the integer part in the input image signal Da. And subtracting or adding the value of the correction bit data Cd from the input image signal Da, adding the correction bit data, converting the pixel value, and outputting an image signal with an expanded number of gradations (bit number) doing. Therefore, according to the image processing apparatus or the image processing method of the first embodiment, even when the input image signal Da includes a noise component, the gradation change of the edge of the image and the gradation of the edge of the image that appears next to it are displayed. Since the gradation change due to the noise component existing between the changes can be determined as the gradation change due to the noise component, the width between the non-noise gradation change position and the non-noise gradation change position that appears next Can be accurately detected and can be distinguished from the gradation change due to the noise component, so that the inclination data value Ksl can be calculated without being influenced by the noise component. Further, according to the image processing apparatus or the image processing method of the first embodiment, the correction value that sequentially changes with the gradation value by the inclination data value Ksl can be obtained without significantly increasing the circuit scale, and the gradation value is gradually reduced. By obtaining appropriate correction bit data Cd in the pixel region where the angle changes and the non-noise gradation change position, the gradation change of the image signal can be converted more smoothly and the number of bits can be expanded. In addition, since the obtained correction bit data Cd is added to the gradation value of the input image signal Da to extend the number of gradations, the number of gradations can be processed in real time, and the gradation can be maintained while leaving a slight change due to noise components. Changes can be converted more smoothly.

  As described above, according to the image processing apparatus or the image processing method of the first embodiment, even when a noise component is included in the input image signal Da, the gradation is gradually increased without significantly increasing the circuit scale in real time. Correction bit data Cd obtained by more appropriately detecting a gradation change can be obtained without erroneous detection due to a noise component in an image region whose value changes. For this reason, the number of gradations of the image signal is expanded by the correction bit data, and the image details are not lost or the sharpness is not lost in the gradation-converted output image signal Db. Alternatively, the pseudo contour is reduced, the gradation change is smooth, and an image signal with an extended number of gradations can be obtained.

<< 1-5 >> Modifications.
In the above description, the pixel area determination unit 20 determines each pixel as a pixel belonging to a flat gradation area, a pixel belonging to a noise area, or a pixel at a non-noise gradation change position based on the magnitude of the gradation change amount df. In the above description, the case where three types of region signals Ar = 0, 1, and 2 are generated as region determination results has been described. The flat / noise region width determination unit 30 using the region signal Ar in FIG. The inclination detection unit 40 and the correction bit generation unit 50 switch processing in two ways: a non-noise gradation change position (Ar = 2), a gradation flat region, and a noise region (Ar = 0, 1) (FIG. 7, described with reference to FIGS. 13, 15, and 16), it is also possible to set the region determination result as a region signal Ar <b> 1 that is expressed in binary. For example, based on the determination result that each pixel is classified as one of the pixel belonging to the flat gradation region, the pixel belonging to the noise region, or the pixel at the non-noise gradation change position, the region signal is output in the flat gradation region and the noise region. Ar1 = 0 and the region signal Ar1 = 1 at the non-noise gradation change position, and processing in the flat / noise region width determination unit 30, the inclination detection unit 40, and the correction bit generation unit 50 is performed by the region signal Ar1. You may make it switch. Also in this case, it can be configured similarly to the case of FIG. 1, and the same effect can be obtained.

  In the above description, the gradation change amount when the gradation conversion means 1 detects the gradation change of the input image signal Da is the difference of the least significant 1 bit, that is, the gradation value between pixels. A case where the difference is +1 and −1 and the gradation change amount limiting unit 12 in the gradation change amount detecting unit 10 restricts the value of the gradation change amount df to be within a range from −1 to +1. Although described above, the present invention is not limited to such a case. For example, when a change in the range between the maximum value +2 and the minimum value −2 of the gradation value difference between pixels is obtained, a predetermined maximum value is obtained. When the gradation value changes within a gradation difference of a value larger than one gradation of the image signal by limiting the value of the gradation change amount df based on (positive value) and minimum value (negative value) In addition, the number of gradations can be extended. The range for limiting the value of the gradation change amount df is a value for determining whether the image area is a smooth change by gradation conversion, whether it is retained as it is, or whether it is corrected as a gradation change that becomes a gradation jump. An appropriate value may be set from the characteristics of the input image signal, the level at which the gradation jump occurs, and the like. The configuration and operation in this case are the same as the configuration and operation shown in FIG. 1 except for the range of the difference in gradation value between pixels for which a change in gradation value is to be obtained. Here, when the change in the range of the maximum difference +2 and the minimum value −2 of the gradation value difference between the pixels is obtained, the range of restriction in the gradation change amount limiting unit 12 in the gradation change amount detecting unit 10, That is, the value of the amplitude limit gradation change amount Xlm is a value in the range from +2 to -2, and the gradient of the gradation value is also a value in the range from -2 to +2.

  In the above description, the gradation change amount limiting means 12 in the gradation change detection means 10 changes the value of the gradation change amount df from −1 to +1 (for the least significant 1 bit of the gradation value of the input image signal Da). The image processing apparatus and the image processing method for limiting to be within the range of (difference)) have been described. However, for example, the gradation change amount limiting means 12 in the gradation change amount detection means 10 determines the value of the gradation change amount df from a predetermined maximum value (a positive value greater than one gradation). You may restrict | limit so that it may be in the range to the minimum value (Absolute value is a negative value larger than 1 gradation). FIG. 20 is a diagram for explaining another example of the operation of the pixel value calculation unit shown in FIG. 1, and the input image signal Da (or the delay-compensated image signal Dd from the delay compensation unit 91), The relationship between the correction bit data Cd from the correction bit generation means 50 and the output image signal Db output from the pixel value calculation means 92 is shown. In the example of FIG. 20, the gradation converting means 1 shown in FIG. 1 detects the change in the gradation value of the input image signal Da in the horizontal direction up to the difference of gradation values “2”, and the level between pixels is detected. A change within the range from the maximum value +2 to the minimum value −2 of the difference between the tone values is obtained. Then, as shown in FIG. 20, a correction value (correction bit data) is generated for an 8-bit input image signal Da by an additional bit E of a value indicated by 5 bits by an integer part 1 bit and a decimal part 4 bits. Then, it is added to the input image signal Da and converted to an output image signal Db with the number of gradations extended to 12 bits. At this time, the pixel value calculation unit 92 sets a value such that the position of the integer part of the additional bit E in the correction bit data Cd matches the position of the integer part of the image signal Dd (the least significant bit corresponds to the integer 1) The image signal Dd is subtracted or added according to the sign, and the output image signal Db is added with a bit string corresponding to the number of bits indicating the decimal part of the input image signal Da. In this case, not only the gradation change of one gradation which is the minimum gradation value of the input image signal Da but also the quantization noise of one gradation or more (for example, due to gradation jump) in the input image signal Da. ) Is included, the same effect as in the first embodiment can be obtained.

  Furthermore, in the above description, the input image signal Da to be input is m = 8 bits, the additional bit E in the correction bit data Cd is generated with 4 bits, and the output image signal Db having n = 12 bits is generated. Although the case where the key number is extended is shown, the number of gradations m of the input image signal Da is not limited to this. For example, the number of gradations m of the input image signal Da may be 10 bits. In this case, an output image signal Db of n = 14 bits can be obtained, and a more accurate image signal can be obtained. Even when the number of bits of the correction bit data Cd (that is, the correction value aCd) is increased, the number of bits in the configuration up to the generation of the correction bit data Cd, that is, the flat / noise region width data predicted value dst, It can be configured only by increasing the number of bits in the inclination data value Ksl and the correction value aCd, and the same effect can be obtained.

  In the above description, the edge change detecting means 60 sets the correction restriction coefficient rc that restricts correction when it is determined that the gradation change is caused by the edge, and the pixel where the edge exists (correction restriction coefficient rc = 0). ), The gradation value is not added, and the contour information of the image is maintained. However, the gradation change amount is limited by the gradation change amount limiting means 12 in the gradation change amount detection means 10. When obtaining the amplitude-limited gradation change amount Xlm whose value is restricted so that the value of df is in the range from −1 to +1, the value of the gradation change amount df is out of the range from −1 to +1. In such a case, even if the value is set to 0 (that is, the amplitude limit gradation change amount Xlm = 0), the correction value aCd after the pixel that becomes the edge can be set to 0 similarly. . That is, if the amplitude limit gradation change amount Xlm = 0 at a pixel whose edge is outside the range from −1 to +1, the gradation change value becomes 0, and the gradient data value Ksl = 0. Further, since the correction value aCd = (rc) × (Xlm) = 0 can be obtained, the same correction value aCd as that when the correction limiting coefficient rc = 0 is obtained.

  Further, in the above description, the gradation converting means 1 detects the gradation change amount in the horizontal direction and generates the correction bit data Cd for the gradation value in the change in the gradation value in the horizontal direction of the input image signal Da. In the above description, the correction bit data Cd is added to the image signal to convert the gradation to reduce the quantization noise. However, the gradation change in the vertical direction of the input image signal Da (the arrangement direction of the screen display scanning lines). By detecting the amount, generating the correction bit data Cd and adding it to the image signal, the change of the gradation value in the vertical direction may be smooth, and the same effect can be obtained. In this case, the gradation change amount detection means 10 calculates the gradation change amount df based on the difference between the target pixel and the adjacent pixel of the target pixel (adjacent pixel on the scanning line one level higher in the vertical direction). The delay processing for one pixel of each signal in the flat / noise area width determination means 30, the inclination detection means 40, and the correction bit generation means 50 is obtained as the delay processing for one line (for one horizontal scanning period). What is necessary is just to comprise so that the change of the gradation value in the pixel arrange | positioned adjacent to the direction may be detected.

  Furthermore, in the above description, it has been described that each configuration of the gradation conversion means 1 in the image processing apparatus is a hardware configuration, but the present invention is configured to be realized by software processing in program control. May be.

<< 2 >> Embodiment 2
<< 2-1 >> Configuration.
FIG. 21 is a block diagram schematically showing an example of the configuration of an image processing apparatus according to the second embodiment of the present invention (that is, an apparatus capable of performing the image processing method according to the second embodiment). In FIG. 21, the same reference numerals are given to the same or corresponding components as those shown in FIG. 1 (Embodiment 1). In the image processing apparatus according to the second embodiment, the configuration and operation of the pixel area determination unit 20b in the gradation conversion unit 2 constituting the image processing apparatus are the same as those in the gradation conversion unit 1 in the first embodiment. This is different from the 20 configurations and operations.

  FIG. 22 is a block diagram schematically showing an example of the configuration of the pixel area determination unit 20b shown in FIG. FIGS. 23A and 23B are diagrams illustrating an example of pixels determined by the pixel region determination unit 20b as noise components in the image processing apparatus according to the second embodiment, and FIG. And a noise component composed of two consecutive “one gradation changing pixels” whose gradation value is increased by one gradation from the immediately following pixel, and FIG. A noise component composed of two consecutive “1 gradation changing pixels” whose tone value is reduced by 1 gradation is shown. 23A and 23B, the horizontal axis indicates the horizontal pixel position, and the vertical axis indicates the gradation value La (x) of the input image signal Da at each pixel position.

  In the image processing apparatus of the first embodiment, the pixel area determination unit 20 in the gradation conversion unit 1 has the configuration shown in FIG. 5, and the gradation value of one '1 gradation change pixel' is one. The gradation change (FIGS. 4A and 4B) that increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel is determined as a noise component. For this reason, in the image processing apparatus of the first embodiment, the value of the gradation change amount df in the target pixel at the pixel position x, the target pixel at the pixel position x, the adjacent pixel at the pixel position x−1 immediately before it, Based on the polarity of the gradation change amount df in each of the adjacent pixels at the pixel position x + 1 immediately after the pixel of interest, each pixel is classified into a pixel belonging to the gradation flat area, a pixel belonging to the noise area, and a non-noise gradation change position. It was determined that the pixel was classified into one of the pixels.

  However, not only a gradation change in which the gradation value of one '1 gradation changing pixel' increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel, A noise component also includes a gradation change in which the gradation value of a plurality of consecutive “one gradation changing pixels” increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. In some cases, it may be desirable to determine that the gradation change is caused by. Therefore, in the image processing apparatus of the second embodiment, the gradation value of one “one gradation changing pixel” is increased by one gradation relative to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. In addition to the gradation change that decreases, the gradation value of two consecutive “one gradation change pixels” increases by one gradation relative to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel, or The decreasing gradation change is also determined to be a gradation change (high frequency change) due to a noise component.

  In the image processing apparatus of the second embodiment, the gradation change amount detection means 10 in the gradation conversion means 2 obtains the gradation change amount df between adjacent pixels in the horizontal direction in the input image signal Da, and An amplitude limited gradation change amount Xlm is obtained by limiting the value of the gradation change amount df to a predetermined range.

  The gradation change amount df from the gradation change amount calculation means 11 in the gradation change amount detection means 10 is input to the pixel region determination means 20b. Based on the value of the gradation change amount df, the pixel area determination unit 20b classifies each pixel as one of a pixel belonging to a flat gradation area, a pixel belonging to a noise area, and a pixel at a non-noise gradation change position. The determination is performed, and an area signal Ar2 indicating the area determination result of the pixel is generated and output. The pixel region determination unit 20b outputs the generated region signal Ar2 to the flat / noise region width determination unit 30, the inclination detection unit 40, and the correction bit generation unit 50. The pixel area determination unit 20b is configured such that the gradation value of one “1 gradation changing pixel” is the gradation value of the immediately preceding pixel and the gradation of the immediately following pixel, as shown in FIGS. A gradation change (high-frequency change) that increases or decreases by one gradation with respect to the tone value is determined not only as a noise region, but also two continuous two as shown in FIGS. The tone value of '1 tone change pixel' is also a noise region where tone change (high frequency change) increases or decreases by 1 tone with respect to the tone value of the immediately preceding pixel and the tone value of the immediately following pixel. It is determined.

In FIG. 23 (a), the input image signal in advance the target pixel at the pixel position x ₁ of Da is the gradation variation df = + 1, floor in the pixel of the pixel position x ₁ +1 after one pixel of the pixel of interest The tone change amount df = 0, the tone change amount df = −1 at the pixel position x ₁ +2 two pixels after the target pixel, and the pixel position x ₁ −1 one pixel before the target pixel. The gradation change amount df = 0 in the pixel, and the gradation change amount df = 0 in the pixel at the pixel position x ₁ -2 two pixels before the target pixel. Further, in FIG. 23 (b) showing another example, keep the target pixel at the pixel position x ₂ of the input image signal Da is the gradation variation df = -1, the pixel position after 1 pixel of interest The gradation change amount df = 0 in the pixel x ₂ +1, and the gradation change amount df = + 1 in the pixel position x ₂ +2 two pixels after the target pixel, which is one pixel before the target pixel. The gradation change amount df = 0 in the pixel at the pixel position x ₂ −1, and the gradation change amount df = 0 in the pixel at the pixel position x ₂ -2 two pixels before the target pixel.

As shown in FIGS. 23A and 23B, the gradation value of two consecutive “1 gradation changing pixels” is 1 with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. When there is a gradation change (high-frequency change) that increases or decreases by the gradation, the absolute value of the gradation change amount df is less than a relatively small value Nd, and the polarity of the gradation change amount of the pixel of interest There is a change in which the polarity of the gradation change amount of the adjacent pixel adjacent to the pixel position two pixels before the target pixel or the polarity of the gradation change amount of the adjacent pixel adjacent to the pixel position two pixels after the target pixel exists. To do. That is, in the case of FIG. 23 (a), a tone variation df _(x 1) = + 1 at the pixel the pixel position _{x 1,} floor at the pixel position _x 1 +2 pixel after 2 pixels of the pixel position _{x 1} The tone change amount df (x ₁ +2) = − 1 and the polarity is opposite. In the case of FIG. 23 (b), the gradation variation df _(x 2) = in the pixel the pixel position _{x 2} - 1, at the pixel position _x 2 +2 pixels after two pixels of the pixel position _{x 2} The gradation change amount df (x ₂ +2) = + 1, and the polarity is opposite.

  Therefore, in the case where there is a gradation change in which the gradation value of one '1 gradation changing pixel' increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. The same determination as the polarity determination is possible. In the second embodiment, the gradation value of two consecutive “1 gradation changing pixels” increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. The tone change can be determined by the absolute value | df | of the gradation change amount df and the polarity of the gradation change amount df in each pixel and the pixels located adjacent to each other before and after the pixel.

  The pixel area determination unit 20b according to the second embodiment includes a pixel of interest at the pixel position x, an adjacent pixel at the pixel position x-1 immediately before the pixel, an adjacent pixel at the pixel position x-2 immediately before, and immediately after the pixel of interest. 4 (a) and (b) and FIGS. 23 (a) and (b) from the gradation change amount df in the adjacent pixel at the pixel position x + 1 and the adjacent pixel at the pixel position x + 2 immediately thereafter. A pixel region that is a noise component as shown is determined, and a pixel in a gray level flat region where the gray level change amount df = 0, a gray level flat region, and a non-noise gray level change position other than the noise region. By determining the pixel and classifying each pixel as one of a pixel belonging to a flat gradation region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position, the region determination result of this pixel Generate and output area signal Ar2 indicating That.

  More specifically, the pixel region determination unit 20b first delays the pixel with respect to the gradation change amount df from the gradation change amount detection unit 10, for example, and thereby the gradation change amount df of the target pixel at the pixel position x. (X), the gradation change amount df (x−1) of the pixel at the immediately preceding pixel position x−1 (immediately adjacent pixel), and the pixel at the immediately preceding pixel position x−2 (immediately adjacent pixel). The gradation change amount df (x−2), the gradation change amount df (x + 1) of the pixel at the pixel position x + 1 immediately after the pixel of interest (immediately adjacent pixel), and the pixel at the pixel position x + 2 immediately after it (adjacent immediately adjacent to each other) The gradation change amount df in the range of a total of five pixels of the gradation change amount df (x + 2) of the pixel) is obtained, the magnitude of the gradation change amount df (x) of the target pixel, and the gradation of the target pixel The polarity of the change amount df (x) and the gradation change amount df (x−1) of the adjacent pixel and the adjacent pixel From the relationship with the polarities of df (x−2), df (x + 1), and df (x + 2), each pixel is classified into a pixel belonging to a flat gradation region, a pixel belonging to a noise region, and a pixel at a non-noise gradation change position. Judgment to classify into either.

  A gradation change in which the gradation value of one “1 gradation changing pixel” increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel (FIG. 4A). And (b)), and the tone value of two consecutive “1 tone changing pixels” is the first level relative to the tone value of the immediately preceding pixel and the tone value of the immediately following pixel. In the case of a noise component resulting in a tone change that increases or decreases by a tone (see FIGS. 23A and 23B), the absolute value of the tone change amount df is a relatively small value in a pixel including the noise component. There is a change that is less than or equal to Nd and that has a polarity opposite to either the front or rear of the gradation change amount df in the pixel range.

  Therefore, the pixel region determination unit 20b determines the gradation change amounts df (x) and df (x−1) in the target pixel at the pixel position x, the immediately adjacent pixel, the immediately adjacent pixel, the immediately adjacent pixel, and the immediately adjacent pixel. ), Df (x−2), df (x + 1), and df (x + 2), each pixel is divided into a pixel belonging to the gradation flat region, a pixel belonging to the noise region, and a pixel at the non-noise gradation change position. Determination to classify into any one of the following determinations B1 to B3 is performed, and an area signal Ar2 indicating the determination result is output.

(Decision B1)
When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0, the target pixel is a pixel belonging to the flat gradation region, and the region signal Ar2 is represented by Ar2 = 0.

(Decision B2)
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( x-1) the opposite polarity (when the adjacent pixel has a high frequency component), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( x + 1) the opposite polarity (when the adjacent pixel has a high frequency component), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( x-2) When the polarity is opposite to the polarity (when the adjacent pixel has a high frequency component), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( When the polarity is opposite to the polarity of x + 2) (when the adjacent pixel has a high frequency component),
The target pixel is a pixel belonging to the noise region, and the region signal Ar2 is Ar2 = 1.

(Decision B3)
If it is determined in the determination B1 that the target pixel is not a pixel belonging to the flat gradation area, and it is determined in the determination B2 that the target pixel is not a pixel belonging to the noise area, the target pixel position x The pixel is a pixel at a non-noise gradation change position, and the area signal Ar2 is Ar2 = 2.
The value of the area signal Ar2 is not limited to the above value, and may be other values as long as each area determination result is indicated.

By performing the determination B1 to B3, as shown in FIG. 23 (a), the pixels of the pixel position _{x 1} and the pixel in pixel position _x 1 +2 is determined that the pixels belonging to the noise region, FIG. 23 (b ), The pixel at the pixel position x _{2 and} the pixel at the pixel position x ₂ +2 are determined to belong to the noise region.

  FIG. 22 shows an example of the configuration of the pixel area determination unit 20b that performs the determinations B1 to B3. In FIG. 22, the same reference numerals are given to the same or corresponding components as those shown in FIG. 5 (Embodiment 1). The pixel area determination means 20b shown in FIG. 22 is the pixel area shown in FIG. This is different from the determination unit 20.

  In FIG. 22, the pixel change amount extraction unit 21 b in the pixel area determination unit 20 b includes, for example, four pixel delay units 210 to 213. When the gradation change amount df from the gradation change amount detection means 10 is input to the pixel change amount extraction means 21b, the pixel delay means 210 to 213 in the pixel change amount extraction means 21b convert the gradation change amount df into the pixel. Delayed, the gradation change amount df (x) at the pixel of interest at pixel position x, the gradation change amount df (x-2) at pixel position x-2, and the gradation change amount df (x) at pixel position x-1. -1), the gradation change amount df (x + 1) at the pixel position x + 1 and the gradation change amount df (x + 2) at the pixel position x + 2 are extracted.

  The absolute value level comparison means 22 receives the gradation change amount df (x) at the target pixel at the pixel position x, and uses the absolute value | df (x) | of the gradation change amount as a predetermined threshold value (0 and noise). The absolute value of the gradation change amount df (x) is compared with a relatively small value Nd corresponding to the gradation difference in the component (for example, a value twice the minimum gradation difference, ie, Nd = 2). The comparison result nlf indicating the result is obtained. The adjacent position polarity determination means 23 receives the gradation change amount df (x) and the gradation change amount df (x−1), and the polarity of the gradation change amount df (x) and the gradation change amount df. It is determined whether there is a change (high-frequency component change) in which the polarity of (x-1) is opposite, and a determination result pf1 indicating a binarized value is obtained as a determination result. The adjacent position polarity determination means 24 receives the gradation change amount df (x) and the gradation change amount df (x + 1), and the polarity of the gradation change amount df (x) and the gradation change amount df (x + 1). It is determined whether there is a change in which the polarity is opposite (change in high-frequency component), and a determination result pf2 indicating a binarized value is obtained as a determination result.

  The adjacent position polarity determination unit 27 receives the gradation change amount df (x) and the gradation change amount df (x−2), and the polarity of the gradation change amount df (x) and the gradation change amount df. It is determined whether there is a change (high-frequency component change) in which the polarity of (x-2) is opposite, and a determination result pf3 indicating a binarized value is obtained as a determination result. The adjacent position polarity determination unit 28 receives the gradation change amount df (x) and the gradation change amount df (x + 2), and the polarity of the gradation change amount df (x) and the gradation change amount df (x + 2). It is determined whether there is a change in which the polarity is opposite (change in high-frequency component), and a determination result pf4 indicating a binarized value is obtained as a determination result.

  The adjacent position polarity determination means 27 is opposite at the pixel position x-2 that is two pixels away from the pixel position x based on the polarity of the gradation change amount df (x) and the gradation change amount df (x-2). It is determined whether there is a change in the high-frequency component that changes the direction, that is, whether the polarity of the gradation change amount df (x) and the gradation change amount df (x-2) are opposite, and the determination result at the pixel position x is A determination result pf3 indicating a binarized value is obtained. The value of the determination result pf3 is obtained similarly to the case of the determination result pf1 in the adjacent position polarity determination unit 23. For example, the polarity of the gradation change amount df (x) and the gradation change amount df (x-2) are opposite. In the case of polarity, the value is “1” (pf3 = 1), and in the case of the same, the value is “0” (pf3 = 0). Note that the value of the determination result pf3 is not limited to this, and may be another value as long as it indicates whether or not the codes do not match.

  Similarly, the adjacent position polarity determination unit 28 uses the polarities of the gradation change amount df (x) and the gradation change amount df (x + 2) at the pixel position x + 2 that is two pixels away from the pixel position x in the opposite direction. Whether the polarity of the gradation change amount df (x) is opposite to the polarity of the gradation change amount df (x + 2). As a determination result at the pixel position x, binary is determined. The determination result pf4 indicating the converted value is obtained. The value of the determination result pf3 is obtained similarly to the case of the determination result pf1 in the adjacent position polarity determination unit 23. For example, the polarity of the gradation change amount df (x) and the gradation change amount df (x-2) are opposite. In the case of polarity, the value is “1” (pf4 = 1), and in the case of the same, the value is “0” (pf4 = 0). Note that the value of the determination result pf4 is not limited to this, and may be another value as long as it indicates whether the codes do not match.

  The combination means 29 receives the determination results pf1 and pf2 from the adjacent position polarity determination means 23 and 24 and the determination results pf3 and pf4 from the adjacent position polarity determination means 27 and 28, and combines the determination results pf1 to pf4. Then, the polarity determination result pof is obtained when the polarity is opposite to the gradation change amount in any of the adjacent pixels and the adjacent pixels with respect to the gradation change amount df (x) at the pixel position x. The synthesizer 29 is configured by, for example, an OR gate, and obtains the polarity determination result pof = 1 when any of the determination results pf1 to pf4 is “1”. As a result, it was determined that the gradation change amount df (x) at the pixel position x is opposite in polarity to any one of the gradation change amount in the adjacent pixel and the adjacent pixel before or after the pixel position x. Results are obtained.

  The region signal generation unit 26 receives the comparison result nlf from the absolute value level comparison unit 22 and the polarity determination result pof from the synthesis unit 29, and determines the pixel region at the target pixel at the pixel position x based on the determinations B1 to B3. Then, it is determined that each pixel is classified into one of the pixel belonging to the flat gradation region, the pixel belonging to the noise region, and the pixel at the non-noise gradation change position, and the region signal Ar2 indicating each region determination result is generated. Then output.

  As described above, the pixel region determination unit 20b determines each pixel based on the gradation change amount df based on the gradation change amount df of the target pixel, the polarity of the target pixel, the polarity of the adjacent pixel, and the polarity of the adjacent pixel. Can be classified into any one of a pixel belonging to a flat gradation region, a pixel belonging to a noise region, and a pixel at a non-noise gradation change position, and a region signal Ar2 indicating the region determination result of the pixel can be obtained. In this pixel area, the gradation change due to the noise component is determined by the magnitude of the gradation change amount df, the polarity of the target pixel, the polarity of the adjacent pixel, and the polarity of the adjacent pixel. Can be determined.

  The pixel change amount extraction unit 21b in FIG. 22 is configured to extract the gradation change amount df with a delay. However, when the gradation change amount detection unit 10 determines the gradation change amount df, It may be obtained by obtaining the gradation change amount of x and its adjacent pixels and adjacent pixels. In this case, the pixel delay in the pixel change amount extraction means 21b can be omitted. In the determinations B1 and B2, the value of the gradation change amount df and the polarity of the gradation change amount at the pixel position x, the adjacent pixel positions x-1, x + 1, and the adjacent pixel positions x-2, x + 2 Therefore, the area to which each pixel belongs is determined, and therefore, a configuration is adopted in which only the polarity is obtained without obtaining the gradation change amounts at the adjacent pixel positions x-1, x + 1 and the adjacent pixel positions x-2, x + 2. Also good.

<< 2-2 >> Operation.
Similar to the area signal Ar in the first embodiment, the area signal Ar2 that is the determination result of the pixel area in the pixel area determination unit 20b is the flat / noise area width determination unit 30, the inclination detection unit 40, and the correction bit generation unit 50. Is output. The flat / noise area width determination means 30, the inclination detection means 40, and the correction bit generation means 50 predict the flat / noise area width after the non-noise gradation change position based on the area signal Ar2. The correction bit data Cd calculated from the data prediction value dst, the gradient of the gradation value (gradient data value Ksl), and the gradient data value Ksl is obtained. Then, the pixel value calculation unit 92 adds the correction bit data Cd to the input image signal Da to convert the pixel value, and generates an output image signal Db with an extended number of bits. These configurations and operations are the same as those in the first embodiment.

  In the image processing apparatus according to the second embodiment, the gradation change amount df is obtained, and on the basis of the gradation change amount, each pixel is classified into a pixel belonging to a flat gradation region, a pixel belonging to a noise region, and a non-noise gradation change position. Is determined, the correction bit data is generated by obtaining the flatness / noise area width and inclination in the gradation change from the area determination result and the gradation change amount df, and the correction bit data is generated in the input image signal Da. The operation of outputting the output image signal Db whose number of gradations is expanded by adding data is the same as the operation in the description of the first embodiment (the description with reference to FIGS. 15 and 16).

  Also, the determination of classifying each pixel as one of a pixel belonging to a flat gradation region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position based on the gradation change amount in the pixel region determination unit 20b. This is the same as the description of the first embodiment (the description with reference to the flowchart of FIG. 17). However, in the second embodiment, the gradation change amount and the polarity of the pixel obtained in steps S120 and S121 in FIG. 17 are located in the five-pixel range of the target pixel at the pixel position x and the two pixels before and after it. The gradation change amounts df (x), df (x−1), df (x + 1), df (x−2), and df (x + 2) in the pixel to be processed, and the processing from step S122 to step S129 for determining the polarity is performed. The second embodiment is different from the first embodiment in that the gradation change amounts df (x−2) and df (x + 2) are also performed.

<< 2-3 >> Effect.
As described above, in the image processing apparatus or the image processing method according to the second embodiment, the gradation value of one “1 gradation changing pixel” is the gradation of the immediately preceding pixel from the gradation change amount df. In addition to the gradation change that increases or decreases by one gradation with respect to the value and the gradation value of the immediately following pixel, the gradation value of two consecutive “1 gradation changing pixels” is the gradation value of the immediately preceding pixel. In addition, each pixel is classified into a pixel belonging to a flat gradation region, a pixel belonging to a noise region, so that a gradation change that increases or decreases by one gradation with respect to the gradation value of the immediately following pixel is also determined as a noise component. It is determined that the pixel is classified into one of the non-noise gradation change positions, the flat / noise area width data predicted value dst is obtained based on the area signal Ar2 indicating the determination result, and the non-noise gradation change position (Ar2 = 2) Amplitude-limited gradation change amount Xlm and flat / noise area width data The slope data value Ksl is obtained from the data predicted value dst. Therefore, according to the image processing apparatus or the image processing method of the second embodiment, even when the input image signal Da contains a noise component, the gradation change of the edge of the image and the gradation of the edge of the image appearing next Since the gradation change due to the noise component existing between the change and the gradation change due to the noise component can be more accurately determined, the non-noise gradation change position and the non-noise gradation change position that appears next Since the flat / noise area width indicating the width between them can be accurately detected and can be distinguished from the gradation change due to the noise component, the slope data value Ksl can be calculated without being affected by the noise component. In addition, according to the image processing apparatus or the image processing method of the second embodiment, the correction value that sequentially changes with the gradation value based on the inclination data value Ksl can be obtained without increasing the circuit scale significantly, and the gradation value gradually increases. By obtaining appropriate correction bit data Cd in the pixel region where the angle changes and the non-noise gradation change position, the gradation change of the image signal can be converted more smoothly and the number of bits can be expanded. In addition, since the obtained correction bit data Cd is added to the gradation value of the input image signal Da to extend the number of gradations, the number of gradations can be processed in real time, and the gradation can be maintained while leaving a slight change due to noise components. Changes can be converted more smoothly.

  As described above, according to the image processing apparatus or the image processing method of the second embodiment, even when a noise component is included in the input image signal Da, the gradation is gradually increased without significantly increasing the circuit scale in real time. Correction bit data Cd obtained by more appropriately detecting a gradation change can be obtained without erroneous detection due to a noise component in an image region whose value changes. For this reason, the number of gradations of the image signal is expanded by the correction bit data, and the image details are not lost or the sharpness is not lost in the gradation-converted output image signal Db. Alternatively, the pseudo contour is reduced, the gradation change is smooth, and an image signal with an extended number of gradations can be obtained.

<< 2-4 >> Modifications.
In the above description, the pixel area determination unit 20b determines whether each pixel is a pixel belonging to the flat gradation area, a pixel belonging to the noise area, or a pixel at the non-noise gradation change position based on the magnitude of the gradation change amount df. Although the case where the classification is performed and three types of region signals Ar2 = 0, 1, and 2 are generated as region determination results has been described, the flat / noise region width determination unit 30 using the region signal Ar2 in FIG. Since the inclination detection unit 40 and the correction bit generation unit 50 switch processing in two ways: a non-noise gradation change position (Ar2 = 2), a gradation flat area, and a noise area (Ar2 = 0, 1). It is also possible to use a region signal Ar1 indicating the region determination result as a binary value. For example, based on the determination result that each pixel is classified as one of the pixel belonging to the flat gradation region, the pixel belonging to the noise region, or the pixel at the non-noise gradation change position, the region signal is output in the flat gradation region and the noise region. Ar1 = 0 and the region signal Ar1 = 1 at the non-noise gradation change position, and processing in the flat / noise region width determination unit 30, the inclination detection unit 40, and the correction bit generation unit 50 is performed by the region signal Ar1. You may make it switch. Also in this case, it can be configured in the same manner and can provide the same effect.

  In the above description, the gradation converting means 2 detects the amount of gradation change in the horizontal direction and generates correction bit data Cd for the gradation value in the change of the gradation value in the horizontal direction of the input image signal Da. In the above description, the correction bit data Cd is added to the image signal to convert the gradation to reduce quantization noise. However, the gradation change in the vertical direction of the input image signal Da (the arrangement direction of the screen display scanning lines) By detecting the amount, generating the correction bit data Cd and adding it to the image signal, the change of the gradation value in the vertical direction may be smooth, and the same effect can be obtained. In this case, the gradation change amount detection means 10 calculates the gradation change amount df based on the difference between the target pixel and the adjacent pixel of the target pixel (adjacent pixel on the scanning line one level higher in the vertical direction). The delay processing for one pixel of each signal in the flat / noise area width determination means 30, the inclination detection means 40, and the correction bit generation means 50 is obtained as the delay processing for one line (for one horizontal scanning period). What is necessary is just to comprise so that the change of the gradation value in the pixel arrange | positioned adjacent to the direction may be detected.

  In the second embodiment, points other than those described above are the same as those in the first embodiment.

<< 3 >> Embodiment 3
<< 3-1 >> Configuration.
FIG. 24 is a block diagram schematically showing an example of the configuration of an image processing apparatus according to the third embodiment of the present invention (that is, an apparatus capable of performing the image processing method according to the third embodiment). 24, the same reference numerals are given to the same or corresponding components as those shown in FIG. 1 (Embodiment 1). The image processing apparatus according to the third embodiment includes the gradation conversion unit 3, and the configuration and operation of the pixel region determination unit 70 in the gradation conversion unit 3 are the same as the pixel region in the gradation conversion unit 1 in the first embodiment. This is different from the configuration and operation of the determination means 20.

  As already described, the change in gradation value due to the noise component is a high-frequency change in which the gradation value increases or decreases in a narrow region of several pixels, and the gradation difference has a relatively small amplitude. In the first and second embodiments, one '1st floor is calculated from the value of the gradation change amount df of the target pixel and the polarity of the gradation change amount within the range of the pixel adjacent to or adjacent to the target pixel. This is a high-frequency change in which the tone value of the tone change pixel '(FIGS. 4A and 4B) and two consecutive' 1 tone change pixels' (FIGS. 23A and 23B) increase or decrease. It was determined to be a noise component. Here, as shown in FIGS. 4 (a) and 4 (b) and FIGS. 23 (a) and 23 (b), a high-frequency change in which the gradation value increases or decreases in the range of several pixels is in the range of several pixels. There is a change in the opposite direction of increase and decrease of the gradation value. Therefore, since the low frequency component of the gradation change amount is obtained by averaging (smoothing) the gradation change amount in the range of several pixels, the output of the smoothing process includes the high frequency component in the pixel where the noise component exists. It is removed and approaches 0. On the other hand, if the gradation value changes to some extent and the adjacent pixel is flat, or if the neighboring change is a change in the gradation value in the same direction, the output of the smoothing process is greater than 0, A value corresponding to the magnitude of the absolute value of the tone change amount is obtained. Further, since it is a low-frequency component of the gradation change amount, the polarity (positive or negative sign) of the result of the smoothing process is the same as the polarity of the original gradation change amount.

  Incidentally, the results of the smoothing process to the tone variation df, it is possible to use, instead of the amplitude limiting gradation variation Xlm used in tilt detection unit 40 or the like, in this case, the gradation variation Although the noise component in df is removed, the value of the gradation change amount after the smoothing process changes. The gradual change in the detected gradation value is a relatively small value that detects the minimum gradation value or gradation jump of the input image signal that results in quantization noise or pseudo contour. The change in amplitude leads to erroneous detection when the change in the gradation value is regarded as flat or when it is recognized as a change that smooths the contour range. Therefore, in the third embodiment, the result of the smoothing process on the gradation change amount df is determined based on the presence / absence of a noise component, and each pixel is classified into a pixel belonging to the flat gradation area, a pixel belonging to the noise area, This is used for the determination to classify the pixel as one of the non-noise gradation change positions.

  In the third embodiment, the gradation value of one “1 gradation changing pixel” is set immediately before by the value obtained by smoothing (filtering or adding) the gradation change amount df in the adjacent pixel range. A noise component that causes a gradation change that increases or decreases by one gradation with respect to the gradation value of the pixel and the gradation value of the immediately following pixel is determined, and each pixel is classified into a pixel that belongs to a flat gradation area, a noise area Is determined to be classified into any of the pixels belonging to the non-noise gradation change position, and the flat / noise area width determining means 30, the inclination detecting means 40, and the correction bit generation are performed based on the area signal which is the pixel area determination result. The processing in means 50 is performed.

  The gradation conversion means 3 of the image processing apparatus according to the third embodiment is replaced with the pixel area determination means 20 shown in FIG. 5 in the configuration of the gradation conversion means 1 (see FIG. 1) in the image processing apparatus according to the first embodiment. Then, using the pixel area determination means 70 shown in FIG. 25, a determination is made to classify each pixel as either a pixel belonging to a flat gradation area, a pixel belonging to a noise area, or a pixel at a non-noise gradation change position, An area signal Ar3 is generated and output as a pixel area determination result. The image processing apparatus according to the third embodiment is the same as the image processing apparatus according to the first embodiment except for the configuration and operation of the pixel region determination unit 70.

  The pixel region determination means 70 is a gradation in which the gradation value of one “1 gradation changing pixel” is increased or decreased by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. Changes (changes in high frequency) (for example, FIGS. 4A and 4B) are determined as noise components. Based on the value of the gradation change amount df from the gradation change amount calculation means 11, the pixel area determination means 70 determines each pixel as a pixel belonging to the gradation flat area, a pixel belonging to the noise area, and the non-noise gradation change position. It is determined that the pixel is classified into one of the pixels, and an area signal Ar3 indicating the area determination result of the pixel is generated and output. This area signal Ar3 is output to the flat / noise area width determining means 30, the inclination detecting means 40, and the correction bit generating means 50.

  When there is a gradation change in which the gradation value of one “one gradation change pixel” increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel, Since the gradation difference of the tone change is a relatively small value Nd (for example, Nd = 2) or less, the change in the pixel due to the noise component among the pixels where the absolute value | df |> 0 of the gradation change amount df is The gradation change amount is smoothed using the value of the gradation change amount df in each of the target pixel at the pixel position x, the adjacent pixel at the pixel position x−1 immediately before it, and the adjacent pixel at the pixel position x + 1 immediately after the target pixel. It can be determined by the value of the converted low frequency component.

That is, for example, the case of FIG. 4 (a), the increase or decrease of the gradation values is adjacent (change in frequency is) gradation variation in the sum of three pixels of the target pixel at the pixel position x ₁ and its front and rear of the adjacent pixels The result of smoothing by adding the amount of gradation change to df is
df (x ₁ −1) + df (x ₁ ) + df (x ₁ +1) = 0 + (+ 1) + (− 1) = 0
It becomes. Further, for example, in FIG. 4 (b), the relative gradation change amount df at a total of three pixels of the target pixel of the pixel position x ₂ and its front and rear of the adjacent pixels, a result of the smoothing processing is
df (x ₂ −1) + df (x ₂ ) + df (x ₂ +1) = 0 + (− 1) + (+ 1) = 0
It becomes. As described above, when there is a high frequency change in the pixel due to the noise component, the absolute value of the gradation change amount df is less than or equal to a relatively small value Nd, and the gradation change amount is set in the range of each pixel and the preceding and subsequent pixels. The result of performing the smoothing process to be added is 0 or a value close to 0 with the high frequency component removed.

In addition, in the case of a high frequency change such as the pixel position x = 35 (= x ₂ ) and the pixels before and after the pixel position x = 35 (= x ₂ ) in FIGS. The gradation change amount df at x = 33, 34, 35 (= x ₂ ), that is, the result of smoothing the df (33), df (34), df (35) is
df (33) + df (34) + df (35) = 0 + (+ 1) + (-2) =-1
It becomes. At this time, the absolute value of the smoothing process result is greater than 0, but the polarity (positive or negative sign) of the smoothing process result is opposite to the original gradation change amount df (34) = + 1. This means that there is a change in the reverse direction of increase and decrease within the range of the smoothed pixel.

  Therefore, smoothing processing (for example, addition processing) is performed on the gradation change amount df at the target pixel position and the adjacent pixel position, and the result value and the gradation change amount df at the target pixel position are used as FIG. ) And (b), it is possible to determine the pixel region as the noise component, the pixel in the gradation flat region where the gradation change amount df = 0, and the other non-noise gradation change position.

  FIG. 25 is a block diagram schematically showing an example of the configuration of the pixel area determination means 70 shown in FIG. 25, the same reference numerals are given to the same or corresponding components as those shown in FIG. 5 (Embodiment 1). A pixel region determination unit 70 shown in FIG. 25 includes the pixel change amount extraction unit 21 and the absolute value level comparison unit 22 shown in FIG. 5, and further includes a smoothing unit 72, a polarity determination unit 73, and a comparison unit 74. And an area signal generating means 75.

  In FIG. 25, the pixel change amount extraction means 21 in the pixel area determination means 70 receives the gradation change amount df from the gradation change amount detection means 10 and delays the gradation change amount df by a pixel to obtain the pixel position. The gradation change amount df (x) and the gradation change amount df (x−1) in a total of three pixel ranges centered on the pixel position x based on the pixel of interest x and adjacent pixel positions x−1 and x + 1 positioned before and after the pixel of interest. ) And the gradation change amount df (x + 1). The absolute value level comparison unit 22 receives the gradation change amount df (x) in the target pixel at the pixel position x, and has a predetermined threshold 0 and a relatively small value Nd (for example, the minimum value corresponding to the gradation difference in the noise component). The absolute value level of the gradation change amount df (x) is determined, and a comparison result nlf indicating the result is obtained. The configuration up to this point is the same as that in FIG.

The smoothing means 72 receives the gradation change amount df in the three-pixel range centered on the pixel position x. That is, the gradation change amounts df (x), df (x−1), and df (x + 1) obtained by the pixel change amount extraction unit 21 are input to the smoothing unit 72. The smoothing means 72 has a gradation change amount df (x) at the pixel position x, a gradation change amount df (x−1) at the pixel position x−1, and a gradation change amount df (x + 1) at the pixel position x + 1. And the gradation change amount is smoothed. That is, the smoothing means 72 includes a target pixel at the pixel position x, an adjacent pixel at the pixel position x−1 adjacent immediately before the pixel position x, and an adjacent pixel at the pixel position x + 1 adjacent immediately after the pixel position x. Then, the gradation change amounts df (x−1), d (x), and df (x + 1) of three consecutive pixels are added to perform the gradation change amount smoothing process. The smoothing process by the smoothing means 72 is a process for obtaining a smoothing result (added value) ct0 by the following equation (6).
ct0 = df (x-1) + df (x) + df (x) (6)

In addition, in Formula (6), the case where the addition value of the gradation change amount in three consecutive pixels is set as the smoothing result ct0 is shown, but the addition value (the right side of Formula (6)) is multiplied by 1/3. An average value of gradation change amounts may be obtained as a smoothing result (divided by the number of added gradation change amounts). In consideration easiness of construction and operation processing of the hardware, the other to the addition value of the gradation variation value, for example, 1/4 = 1/2 ² multiplied (in addition gradation variation A smoothing result may be obtained by dividing by the factorial of 2 which is close to the number.

In addition, weighted addition is performed so that the weight of the gradation value at the pixel near the pixel position x, which is the center position, increases, and the weight of the gradation value at a pixel (adjacent pixel) located away from the center position decreases. Thus, a smoothing process may be performed to obtain a smoothing result. For example, the gradation value weight at the pixel position x is set to 1, and the gradation value weight at the adjacent pixel is set to ½, and the smoothing result ct0 is calculated according to the following expression (7) instead of the expression (6). You can also
ct0 = df (x-1) / 2 + df (x) + df (x + 1) / 2 Formula (7)

  The smoothing result ct0 calculated according to the above formula (6) or formula (7) is a value obtained by adding the gradation change amount df of the three pixels as it is or with weighting. For this reason, if there is a gradation change amount in the reverse direction of increase and decrease within the range of 3 pixels, that is, a high-frequency change that is considered to be a noise component, the absolute value of the smoothing result ct0 is 0 or a value close to 0 become. If the absolute value of the smoothing result ct0 is not 0, the polarity of the smoothing result ct0 is opposite to the polarity of the original gradation change amount. When the change in the gradation value within the range of 3 pixels is a change in one direction and no noise component is included, the absolute value of the smoothing result ct0 is greater than 0, and the absolute value of the gradation change amount is A value corresponding to the magnitude is obtained. In this case, the polarity of the smoothing result ct0 is the same as the polarity of the gradation change amount df (x). This is because the smoothing result ct0 is obtained by adding the gradation change amount df in the range of 3 pixels, that is, obtained as a low frequency component of the gradation change amount.

  The polarity determination means 73 includes the gradation change amount df (x) at the target pixel at the pixel position x obtained by the pixel change amount extraction means 21 and the gradation change amount smoothing result ct0 from the smoothing means 72. Entered. The polarity determination unit 73 determines whether the polarity of the gradation change amount df (x) and the polarity of the smoothing result ct0 are opposite to each other, and uses a determination result pl0 indicating a binarized value as the polarity determination result. Ask. The value of the determination result pl0 is, for example, a value “1” (pl0 = 1) when the polarities are opposite to each other, and a value “0” (pl0 = 0) when the polarities are the same. Note that the value of the determination result pl0 is not limited to this, and may be another value as long as it indicates whether the polarities do not match. Thereby, the determination result of the polarities of the smoothing result ct0 and the gradation change amount df (x) is obtained.

  The smoothing result ct0 of the gradation change amount from the smoothing means 72 is input to the comparison means 74. The comparison means 74 compares the level of the absolute value of the smoothing result ct0 with a predetermined threshold value. That is, the comparison unit 74 determines a case where the absolute value | ct0 | ≦ cth0 (ct0 = 0) of the smoothing result is set as a predetermined threshold cth0 = 0 (set to 0 or a value close to 0) for comparison. Then, a comparison result cf0 indicating the result is obtained.

  The value of the comparison result cf0 obtained by the comparison means 74 is binary, for example, the absolute value of the smoothed result | ct0 | ≦ cth0, the comparison result cf0 = 1, and otherwise the comparison result cf0 = 0. Note that the value of the comparison result cf0 is not limited to this, and may be another value as long as the result of comparing the magnitudes of the smoothing results is shown.

  Here, since the smoothing result ct0 is obtained as a low frequency component of the gradation change amount df, if there is a high frequency change in the pixel due to a noise component, the value is 0 or close to 0 excluding the high frequency component. Value. Therefore, for the predetermined threshold value cth0, 0 or a value close to 0 is set in order to determine whether or not the gradation change amount is due to a noise component. In FIG. 25, the smoothing result ct0 is obtained by using the smoothing in the smoothing means 72 as Expression (6), and the predetermined threshold cth0 = 0 is set. The threshold cth0 is the minimum in the image signal. Setting of gradation difference and gradation difference detected when gradation conversion is performed, amplitude of noise / error component included in image signal, weighting in smoothing process in smoothing means 72 (formula (7) ) Etc.) etc. are set.

  The region signal generation means 75 receives the comparison result nlf from the absolute value level comparison means 22, the determination result pl0 from the polarity determination means 73, and the comparison result cf0 from the comparison means 74. The area signal generation means 75 performs a determination to classify each pixel into one of a pixel belonging to a flat gradation area, a pixel belonging to a noise area, and a pixel at a non-noise gradation change position, and an area indicating each area determination result A signal Ar3 is generated and output. That is, the area signal generation means 75 performs the following three determinations C1 to C3, and each pixel is selected from a pixel belonging to a flat gradation area, a pixel belonging to a noise area, and a pixel at a non-noise gradation change position. And the region signal Ar3 is generated.

(Decision C1)
When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0, the comparison result nlf from the absolute value level comparison unit 22 is nlf = 1, and the target pixel is , The pixel belonging to the flat gradation region, and the region signal Ar3 becomes Ar3 = 0.

(Decision C2)
When the gradation change amount df (x) in the pixel of interest at the pixel position x is 0 <| df (x) | ≦ Nd and the determination result pl0 from the polarity determination unit 73 is pl0 = 1 ( The polarity of the smoothing result ct0 and the polarity of the gradation change amount df (x) are opposite to each other), or
When the gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd and the comparison result cf0 from the comparison unit 74 is cf0 = 1 (smooth) The absolute value of the conversion result | ct0 | ≦ 0) is
The target pixel is a pixel belonging to the noise region, and the region signal Ar3 becomes Ar3 = 1.

(Decision C3)
If it is determined in the determination C1 that the target pixel is not a pixel belonging to the flat gradation area, and it is determined in the determination C2 that the target pixel is not a pixel belonging to the noise area, the target position of the pixel position x The pixel is a pixel at a non-noise gradation change position, and the area signal Ar3 is Ar3 = 2.

  As described above, the noise component in which the absolute value of the gradation change amount df is less than or equal to the relatively small value Nd and the reverse change of the increase and decrease exists in the smoothed three-pixel range is determined. Therefore, an area signal Ar3 indicating the result of performing the determination to classify each pixel into one of a pixel belonging to a flat gradation area, a pixel belonging to a noise area, and a pixel at a non-noise gradation change position is obtained, It shows which gradation value changes the pixel has in the gradation flat area, noise area, and non-noise gradation change position.

  26A to 26E show pixel area determination means in the third embodiment for the input image signal Da (see FIG. 3) including the image signal noise components shown in FIGS. 6A to 6G. FIG. 7 is a diagram showing a state in which a pixel area is determined by 70. 26A shows the pixel position x in the horizontal direction of the input image signal Da, FIG. 26B shows the gradation value La (x) of the input image signal Da, and FIG. The gradation change amount df based on the difference in the gradation value between adjacent pixels obtained by the gradation change amount detection means 10 is shown. FIG. 4D is obtained by the smoothing means 72 in the pixel area determination means 70. The smoothing result ct0 in the three-pixel range according to the equation (6) is shown, and FIG. 9E shows the region signal Ar3 output from the region signal generating means 75.

For example, in FIGS. 26A to 26E, the gradation change amount df (FIG. 26C) obtained for the input image signal Da including the noise component (FIG. 26B) is represented by the pixel position x. = 16,24,32,40 obtained as the gradation variation df = + 1 in, but the pixel position _x 1 in the middle of the pixels in each gray-scale flat area, _{x 2} and in the noise component of the pixel at the front and rear, The gradation change amount df is obtained as a non-zero value. When the gradation change amount is input to the pixel area determining means 70, the gradation variation df (x ₁₎ at the pixel position x ₁ is
df (x ₁ ) = + 1 ≦ Nd
In and, df gradation variation in the adjacent pixels _{(x 1) df (x 1} -1), df (x 1 +1) is
df (x ₁ -1) = 0,
df (x ₁ +1) = − 1 <0
Therefore, the smoothing result ct0 obtained by the smoothing means 72 is
ct0 = 0 + (+ 1) + (− 1) = 0 ≦ cth0 (= 0)
Thus, the comparison result cf0 = 1. Therefore, the region signal Ar3 = 1 from the region signal generation means 75 is obtained, which becomes a noise region. Similarly, the region signal Ar = 1 at the pixel position x ₁ +1, which is a noise region.

Further, the gradation change amount df (x ₂ −1) at the pixel position x ₂ −1 in FIGS.
df (x ₂ −1) = + 1 ≦ Nd
The gradation change amounts df (x ₂ -2) and df (x ₂ ) in the adjacent pixel of the pixel at the pixel position x ₂ −1 are
df (x ₂ −2) = 0
df (x ₂ ) = − 2 <0
And the smoothing result ct0 obtained is
ct0 = 0 + (+ 1) + (-2) =-1
It becomes. In this case, | ct0 |> cth0 (= 0) the comparison result cf0 = is a 0, the polarity of the smoothed results CT0 = -1 and tone variation _{df (x 2 -1) = +} 1 is the opposite, the determination The result is pl0 = 1. Therefore, the region signal Ar3 = 1 from the region signal generation means 75 is obtained, which becomes a noise region. Similarly, the region signal Ar = 1 at the pixel positions x ₂ and x ₂ +1, and this becomes a noise region.

  As described above, the pixel area determination unit 70 performs the smoothing process on the gradation change amount df in the three-pixel range including adjacent pixels from the gradation change amounts df (x), df (x−1), and df (x + 1). Depending on the result and the value of the gradation change amount df (x), each pixel in the pixel of interest at the pixel position x is selected from any of the pixels belonging to the flat gradation area, the pixels belonging to the noise area, and the pixels at the non-noise gradation change position. It is possible to obtain a region signal Ar3 indicating the pixel region determination result. As in the first and second embodiments, this pixel area is not determined by detecting the gradation difference within the same gradation value or within a certain value, but the gradation change due to the noise component is represented by the gradation change amount df. And the result of smoothing processing, the influence of noise is small, and classification can be performed more accurately only by the amount of gradation change in adjacent pixels.

  The pixel change amount extraction unit 70 in FIG. 25 is configured to extract the gradation change amount df with a delay. However, when the gradation change amount detection unit 10 obtains the gradation change amount df, the pixel position is extracted. It may be obtained by obtaining the gradation change amount in x and its adjacent pixels and adjacent pixels. In this case, the pixel delay in the pixel change amount extracting means 70 can be omitted.

<< 3-2 >> Operation.
Similar to the area signal Ar (in the case of the first embodiment), the area signal Ar3 that is the determination result of the pixel area by the pixel area determining means 70 is the flat / noise area width determining means 30, the inclination detecting means 40, and the correction bit. It is output to the generation means 50. The flat / noise area width determination means 30, the inclination detection means 40, and the correction bit generation means 50 predict the flat / noise area width after the non-noise gradation change position based on the area signal Ar3. The correction bit data Cd calculated from the data prediction value dst, the gradient of the gradation value (gradient data value Ksl), and the gradient data value Ksl is obtained. Then, the pixel value calculation unit 92 converts the gradation value by adding the correction bit data Cd to the input image signal Da, and extends the number of bits of the image signal. These configurations and operations are the same as those in the first embodiment.

  In the image processing apparatus according to the third embodiment, the gradation change amount df is obtained, and on the basis of the gradation change amount, each pixel is classified into a pixel belonging to a gradation flat region, a pixel belonging to a noise region, and a non-noise gradation change position The correction bit data is generated by obtaining the flatness / noise area width and inclination in the gradation change from the area determination result and the gradation change amount df, and the number of gradations is expanded. The operation of outputting the output image signal Db is the same as that described with reference to FIGS. 15 and 16 regarding the first embodiment. However, a determination process for classifying each pixel into one of a pixel belonging to a flat gradation region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position based on the gradation change amount in the pixel region determination means 70 is performed. This is different from the case of the first embodiment.

  FIG. 27 is a flowchart showing the operation of the pixel area determination means 70 shown in FIG. In FIG. 27, steps that are the same as or correspond to those shown in FIG. 17 (Embodiment 1) are assigned the same reference numerals. In FIG. 27, when the gradation change amount df is input to the pixel region determination means 70, the pixel of interest at the pixel position x and the pixel position x− immediately before it are processed by processing such as pixel delay with respect to the gradation change amount df. The gradation change amount df in each of the one pixel (adjacent pixel) and the pixel (adjacent pixel) at the pixel position x + 1 immediately after the target pixel is expressed as gradation change amounts df (x), df (x−1), df ( x + 1) (step S120).

  The smoothing means 72 adds the gradation change amounts df (x), df (x−1), and (x + 1) in the three-pixel range with the pixel position x as the center, and performs the smoothing process according to the above equation (6). Then, the smoothing result ct0 is obtained (step S140). The smoothing result ct0 is obtained as a low-frequency component of the gradation change amount, and it is determined whether or not there is a gradation change amount in the opposite direction of increase and decrease, that is, a noise component that is a high-frequency change within the range of three pixels. it can.

  Next, the comparing means 74 compares the level of the absolute value of the smoothed result ct0 with a predetermined threshold value cth0 = 0 (sets 0 or a value close to 0) (step S141). When the absolute value | ct0 | ≦ cth0 (ct0 = 0) of the smoothing result (step S142), the comparison unit 74 outputs the comparison result cf0 = 1 (step S143). Otherwise, the comparison result cf0 = 0 is output (step S144). Since the smoothing result ct0 is a low-frequency component of the gradation change amount df, if there is a high-frequency change in the pixel due to the noise component, the value becomes a value close to 0 with the high-frequency component removed, and the comparison result cf0 = 1 shows that there is a change that becomes a noise component.

  Further, the polarity determination means 73 performs a comparison determination between the polarity of the gradation change df (x) in the target pixel at the pixel position x and its polarity (step S151). That is, the polarity determination unit 73 determines whether the gradation change amount df (x) and the polarity of the smoothing result ct0 are opposite (step S152), and obtains the determination result pl0 as the polarity determination result. The polarity determination means 73 outputs pl0 = 1 when the polarities are opposite to each other (step S153), and outputs pl0 = 0 when the polarities are the same (step S154).

  Then, the level of the absolute value of the gradation change amount df (x) in the target pixel at the pixel position x is compared with a predetermined threshold value. In the operation shown in steps S130, S131 and S161 to S164 as follows using the comparison result cf0 and the polarity determination result pl0 by the smoothing result ct0. An area indicating each area determination result by performing a determination to classify each pixel in the target pixel at the pixel position x into a pixel belonging to a flat gradation area, a pixel belonging to a noise area, or a pixel at a non-noise gradation change position. A signal Ar3 is generated. Steps S130 and S131 are the same processing as in FIG.

(Decision C1)
When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0 (step S130), the comparison result nlf from the absolute value level comparison unit 22 is nlf = 1. The target pixel is a pixel belonging to the flat gradation region, and the region signal Ar3 becomes Ar3 = 0 (step S161).

(Decision C2)
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd (step S131), and the determination result pl0 from the polarity determination unit 73 is pl0 = 1. (When the polarity of the smoothing result ct0 is opposite to the polarity of the gradation change amount df (x)) (step S162), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd (step S131), and the comparison result cf0 from the comparison unit 74 is cf0 = 1. If there is (when the absolute value of the smoothing result | ct0 | ≦ 0) (step S162),
The target pixel is a pixel belonging to the noise region, and the region signal Ar3 becomes Ar3 = 1 (step S163).

(Decision C3)
When it is determined in the determination C1 that the target pixel is not a pixel belonging to the flat gradation area, and it is determined in the determination C2 that the target pixel is not a pixel belonging to the noise area (steps S131 and S162) The target pixel at the pixel position x is the pixel at the non-noise gradation change position, and the area signal Ar3 becomes Ar3 = 2 (step S164).

  With the operation shown in FIG. 27, it is determined that each pixel at the pixel position x is classified into one of the pixel belonging to the flat gradation region, the pixel belonging to the noise region, and the pixel at the non-noise gradation change position. Is obtained.

<< 3-3 >> Effect.
As described above, in the image processing apparatus or the image processing method according to the third embodiment, the result of smoothing the gradation change amount df in the three-pixel range including adjacent pixels and the gradation change amount df (x ), A gradation change (in which the gradation value of one “one gradation changing pixel” is increased or decreased by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel ( The decision is made to classify each pixel as either a pixel belonging to a flat gradation region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position so that a high-frequency gradation change) is determined as a noise component. The flat / noise area width data predicted value dst after the non-noise gradation change position (Ar3 = 2) is obtained with reference to the area signal Ar3 indicating the determination result, and at the non-noise gradation change position (Ar3 = 2). Amplitude limit gradation change amount Xlm and flat Noise area width data prediction value gradient of the gradation values from the dst is obtained (slope data value Ksl). Therefore, according to the image processing apparatus or the image processing method of the third embodiment, even when the input image signal Da includes a noise component, the non-noise gradation change position and the non-noise gradation change position that appears next to the non-noise gradation change position Since the gradation change due to the noise component existing between the two can be accurately determined as the gradation change due to the noise component, the flat / noise area width can be accurately detected, and the inclination data value Ksl is affected by the noise component. Can be calculated accurately. Further, according to the image processing apparatus or the image processing method of the third embodiment, the correction value that sequentially changes with the gradation value based on the inclination data value Ksl can be obtained without increasing the circuit scale, and the gradation value is gradually increased. By obtaining appropriate correction bit data Cd in the pixel region where the angle changes and the non-noise gradation change position, the gradation change of the image signal can be converted more smoothly and the number of bits can be expanded. In addition, since the obtained correction bit data Cd is added to the gradation value of the input image signal Da to extend the number of gradations, the number of gradations can be processed in real time, and the gradation can be maintained while leaving a slight change due to noise components. Changes can be converted more smoothly.

  As described above, according to the image processing apparatus or the image processing method of the third embodiment, even when a noise component is included in the input image signal Da, the gradation is gradually increased without significantly increasing the circuit scale in real time. Correction bit data Cd obtained by more appropriately detecting a gradation change can be obtained without erroneous detection due to a noise component in an image region whose value changes. For this reason, the number of gradations of the image signal is expanded by the correction bit data, and the image details are not lost or the sharpness is not lost in the gradation-converted output image signal Db. Alternatively, the pseudo contour is reduced, the gradation change is smooth, and an image signal with an extended number of gradations can be obtained.

<3-4> Modifications.
In the above description, the pixel area determination unit 70 increases the gradation value of one “1 gradation changing pixel” by one gradation relative to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. Alternatively, the gradation change (high-frequency change) that decreases (for example, FIGS. 4A and 4B) is determined as a noise component. However, the gradation value of one “1 gradation change pixel” is described. Is not only the gradation change that increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel, but also the gradation values of two consecutive “1 gradation changing pixels”. A gradation change (for example, FIGS. 23A and 23B) that increases or decreases by one gradation relative to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel is also detected as a noise component. It can also be configured.

  FIG. 28 shows a gradation change in which the gradation value of two consecutive “one gradation changing pixels” increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel. It is a block diagram which shows one structural example of the pixel area | region determination means 70b which determines the noise component of as a noise area | region. In FIG. 28, the same reference numerals are given to the same or corresponding components as those shown in FIG. 25 or FIG. The pixel area determination means 70b shown in FIG. 28 uses the pixel change amount extraction means 21 in FIG. 25 to extract the pixel change amount df that extracts the gradation change amount df in a total of five pixel ranges of the pixel position x and the two pixels before and after the pixel position x. The means 21b has a configuration in which the smoothing means 72 for the 3-pixel range is replaced with the smoothing means 76 for the 5-pixel range. The pixel change amount extraction means 21b in FIG. 28 is the same as that in FIG.

  In FIG. 28, the smoothing unit 76 performs a smoothing process by adding the gradation change amount df in a five-pixel range centered on the gradation change amount df (x), thereby obtaining a smoothing result ct1. Get. While the smoothing means 72 is a smoothing process of the gradation change amount df in the three-pixel range centering on the gradation change amount df (x), the reverse direction of the increase and decrease in the wider five-pixel range Can be obtained as the value of the smoothing result ct0, so that the gradation of two consecutive “1 gradation changing pixels” can be obtained. The smoothing result ct1 can be obtained up to a gradation change in which the value increases or decreases by one gradation with respect to the gradation value of the immediately preceding pixel and the gradation value of the immediately following pixel.

  In the above description, the pixel area determination unit 70 determines each pixel as a pixel belonging to the flat gradation area, a pixel belonging to the noise area, and a pixel at the non-noise gradation change position based on the magnitude of the gradation change amount df. In the above description, the case where three types of region signals Ar3 = 0, 1, and 2 are generated as region determination results has been described. However, the flat / noise region width determination unit 30 using the region signal Ar3, the inclination Since the detection means 40 and the correction bit generation means 50 are switched between the non-noise gradation change position (Ar3 = 2), the gradation flat area, and the noise area (Ar3 = 0, 1), the processing is switched. It is also possible to use a region signal Ar1 indicating the region determination result as a binary value. For example, based on the determination result that each pixel is classified as one of the pixel belonging to the flat gradation region, the pixel belonging to the noise region, or the pixel at the non-noise gradation change position, the region signal is output in the flat gradation region and the noise region. Ar1 = 0 and the region signal Ar1 = 1 at the non-noise gradation change position, and processing in the flat / noise region width determination unit 30, the inclination detection unit 40, and the correction bit generation unit 50 is performed by the region signal Ar1. You may make it switch. Also in this case, the configuration can be the same as in the case of FIG. 24, and the same effect can be achieved.

  In the above description, the gradation converting means 3 detects the amount of gradation change in the horizontal direction and generates correction bit data Cd for the gradation value in the change of the gradation value in the horizontal direction of the input image signal Da. In the above description, the correction bit data Cd is added to the image signal to convert the gradation to reduce the quantization noise. However, the gradation change in the vertical direction of the input image signal Da (the arrangement direction of the screen display scanning lines). By detecting the amount, generating the correction bit data Cd and adding it to the image signal, the change of the gradation value in the vertical direction may be smooth, and the same effect can be obtained.

  In the third embodiment, points other than the above are the same as in the first or second embodiment.

<< 4 >> Embodiment 4
<< 4-1 >> Configuration.
In the gradation conversion means 1 of the image processing apparatus of the first embodiment, the area signal Ar is determined based on the gradation change amount df, and the non-noise gradation change position (Ar = 2) based on the area signal Ar. The flat / noise area width data prediction value dst obtained by predicting the width of the flat / noise area immediately after the reference pixel is obtained, and the amplitude limit gradation change amount Xlm at the non-noise gradation change position (Ar = 2) The gradient (gradient data value) Ksl of the gradation value is calculated from the noise region width data predicted value dst. On the other hand, in the gradation conversion means 4 of the image processing apparatus according to the fourth embodiment, the number of pixels (flat / noise area width data) between the area signal Ar and the non-noise gradation change position obtained using the memory. A configuration is employed in which the gradient data value Ks of the gradation value is obtained from the measured value de) and the amplitude limit gradation change amount Xlm. In this case, the gradient data value of the gradation value is detected without erroneous detection due to a noise component. Ks can be obtained.

  FIG. 29 is a block diagram schematically showing an example of the configuration of an image processing apparatus according to Embodiment 4 of the present invention (that is, an apparatus capable of performing the image processing method of Embodiment 4). 29, the same reference numerals are given to the same or corresponding components as those shown in FIG. 1 (Embodiment 1).

  The image processing apparatus according to the fourth embodiment uses the area signal Ar as a reference and the number of pixels (flat / noise area width data measurement value de) between the non-noise gradation change positions (Ar = 2) and the amplitude-limited gradation. The gradient data value Ks of the gradation value is obtained from the change amount Xlm, the correction bit data Ce is generated from the obtained gradient data value Ks, and the gradation of the input image signal Da (denoted by reference sign Dd). An output image signal Db with an expanded number of bits is generated by performing arithmetic processing for adding a value based on the correction bit data Ce to the value. The image processing apparatus according to the fourth embodiment detects a change in the gradation value in a certain direction (for example, the horizontal direction or the vertical direction) of the image signal, and the gradation value gradient data value Ks without being affected by the noise component. Is obtained, gradation conversion processing for expanding the number of bits of the input image signal Da is performed, and an output image signal Db having a higher gradation than the input image signal Da is output.

  As shown in FIG. 29, the image processing apparatus according to the fourth embodiment performs gradation conversion processing for expanding the number of bits by adding a lower bit to an m-bit digital input image signal Da, and n bits There is a gradation converting means 4 for outputting as an output image signal Db of (n> m). The gradation conversion unit 4 includes a gradation change amount detection unit 10 and a pixel area determination unit 20. The gradation change amount detection means 10 and the pixel area determination means 20 shown in FIG. 29 have the same configuration as that in the first embodiment. Further, the gradation converting means 4 includes a flat / noise area width determining means 80 for obtaining the number of pixels between two non-noise gradation changing positions (Ar = 2) as a flat / noise area width data measurement value de, and an end point. It includes a gradation change amount detection means 42 and an inclination calculation means 43, and a fixed direction of the input image signal Da from the flat / noise region width data measurement value de and the amplitude limited gradation change amount Xlm from the gradation change amount detection means 10. The inclination detection means 41 for detecting the inclination data value Ks of the gradation value (for example, in the horizontal direction) and the area signal Ar from the pixel area determination means 20 are delayed and output from the inclination detection means 41 (inclination data value Ks). ) And the area signal delay compensation means 67 for adjusting the phase, the correction bit generation means 81 for generating the correction bit data Ce from the input inclination data value Ks, and the input image signal Da for a predetermined time. Includes a delay compensation means 93 causes delays, and a pixel value calculating means 94 for adding the correction bit data Ce from the correction bit generation unit 81 to the input image signal Da.

  In the fourth embodiment, as in the case of the first embodiment, a change in which the gradation change amount df between pixels is +1 or −1 with respect to the horizontal direction of the input image signal Da is obtained. A description will be given of a configuration in which gradation conversion processing for expanding the number of gradations of the 8-bit input image signal Da in a region where the change is gradual is performed and output as a 12-bit output image signal Db.

  In FIG. 29, the gradation change amount detecting means 10 detects a gradation change amount df in the horizontal direction of the input image signal Da, and a gradation change amount df which is a difference in gradation value between adjacent pixels in the horizontal direction. Then, the amplitude limit gradation change amount Xlm is output in which the value of the gradation change amount df is limited within a predetermined range (for example, within a range from −1 to +1). The amplitude limit gradation change amount Xlm is used as the gradation change height at the non-noise gradation change position when the inclination detection unit 41 calculates the inclination data value Ks of the gradation value in the horizontal direction. Based on the value of the gradation change amount df from the gradation change amount calculation means 11, the pixel area determination unit 20 assigns each pixel to a pixel belonging to the gradation flat region, a pixel belonging to the noise region, and a non-noise gradation change position. It is determined that the pixel is classified into one of the pixels, and an area signal Ar indicating the pixel area determination result is generated and output.

  An area signal Ar from the pixel area determining means 20 is input to the flat / noise area width determining means 80. The flat / noise region width determining unit 80 is based on the region signal Ar from the pixel region determining unit 20 for each horizontal scanning period, and the flat / noise region between two non-noise gradation change positions (Ar = 2). The number of pixels is counted, and a value obtained by adding 1 to the count value is obtained as the flat / noise area width data measurement value de. That is, from the non-noise gradation change position (Ar = 2) with the pixel (reference pixel) at the non-noise gradation change position (Ar = 2) as a reference (starting point), the non-noise gradation change position after that The number of pixels in the flat / noise area up to the pixel (Ar = 2) is counted, and a value obtained by adding 1 to the number of pixels is defined as a flat / noise area width data measurement value de. In this case, since the gradation value of the pixel from the non-noise gradation change position (Ar = 2) at the start point to the non-noise gradation change position (Ar = 2) as the next end point is detected, flat / noise A memory capable of holding data corresponding to the area width is required. However, in the fourth embodiment, it is possible to obtain a measured value of the flat / noise area width (width when calculating the gradient of the gradation value change) instead of the predicted value of the flat / noise area width. .

  Specifically, the flat / noise area width determining unit 80 initializes the flat / noise area width data measurement value de for each horizontal scanning period, and based on the area signal Ar, the non-noise level that becomes the start pixel position. The number of pixels from the tone change position (Ar = 2) to the next non-noise gradation change position (Ar = 2) which is the end pixel position is counted. Then, the flat / noise area width determining means 80 sets a value obtained by adding 1 to the obtained count value as the flat / noise area width data measurement value de, and sets the end point from the non-noise gradation change position as the start point. The flat / noise area width data measurement value de is held and output until the position immediately before the noise gradation change position. This flat / noise area width data measurement value de is output to the inclination calculating means 43 in the inclination detecting means 41, and when calculating the inclination of the gradation value change in the horizontal direction, Distance). Since the flat / noise area width determining means 80 counts the number of pixels between the non-noise gradation change positions (Ar = 2), not only the pixels in the gradation flat area (Ar = 0) but also the noise area (Ar = 1), the number of pixels is counted as a pixel in a flat period. Therefore, the number of pixels between the non-noise gradation change positions (Ar = 2) can be obtained including the noise region (that is, without being affected by the noise component).

  The inclination detection means 41 includes an area signal Ar from the pixel area determination means 20, an amplitude limit gradation change amount Xlm from the gradation change amount restriction means 12 in the gradation change amount detection means 10, and a flat / noise area. The flat / noise area width data measurement value de from the width determining means 80 is input. The inclination detecting means 41 is based on the area signal Ar from the pixel area determining means 20 and the non-noise gradation change position (Ar = 2) which is the next end point with respect to the non-noise gradation change position (Ar = 2) at the start point. Is determined as the height of gradation change (referred to as “endpoint amplitude limit gradation change amount Xe”), and the end point amplitude limit gradation change amount Xe and flat / noise area width data measurement. From the value de, an inclination data value Ks indicating the inclination of the gradation value from the non-noise gradation change position at the start point to the non-noise gradation change position at the end point is calculated and output to the correction bit generation means 81. The end point amplitude limit gradation change amount Xe is the height of the gradation change used in the calculation of the gradient of the gradation value, and is a value obtained by limiting the gradation change amount df within a range from −1 to +1.

  Specifically, first, the end point gradation change amount detection means 42 in the inclination detection means 41 includes an area signal Ar from the pixel area determination means 20 and an amplitude limited gradation change amount from the gradation change amount detection means 10. From Xlm, an amplitude limit gradation change amount Xlm at the next end point non-noise gradation change position (Ar = 2) with respect to the start point non-noise gradation change position (Ar = 2) is detected, and the end point amplitude limit gradation is detected. Obtained as the change amount Xe. The end point amplitude limit tone change amount Xem, which is the amplitude limit tone change amount Xlm at the end point non-noise tone change position (Ar = 2), is calculated when the tone value gradient is calculated in the horizontal direction. It is used as the height, and is output while being held from the non-noise gradation change position at the start point to the position immediately before the non-noise gradation change position as the end point.

Next, the inclination calculating means 43 in the inclination detecting means 41 includes an end point amplitude limit gradation change amount Xe from the end point gradation change amount detecting means 42 and a flat / noise area from the flat / noise area width determining means 80. The width data measurement value de is input. Since the flat / noise area width data measurement value de is the number of pixels from the non-noise gradation change position (Ar = 2) at the start point to the next end non-noise gradation change position, the slope calculation means 43 determines the end point. An inclination data value Ks that is an inclination of the gradation value is calculated from the amplitude limited gradation change amount Xe and the flat / noise area width data measurement value de by the following equation (8).
Ks = Xe / de (8)

  Both the end point amplitude limit gradation change amount Xe and the flat / noise area width data measurement value de are from the non-noise gradation change position that is the start point to immediately before the non-noise gradation change position that is the end point. Since the gradient data value Ks obtained from the above equation (8) is also stored and input so as to indicate the value of the gradation change height and width when calculating the gradient of the gradation value of It is obtained as the gradient of the gradation value in the period from the non-noise gradation change position to the immediately preceding non-noise gradation change position.

  The end point amplitude limit gradation change amount Xe (that is, the amplitude limit gradation change amount Xlm) is limited within a range from −1 to +1. For example, the end point amplitude limit gradation change amount Xe is +1. In the case where the number of pixels between the non-noise gradation change position (Ar = 2) from the start point to the end point position is 7 pixels and the flat / noise area width data measurement value de = 8 is input, it is calculated. The slope data value Ks is 1/8. Further, when the end point amplitude limit gradation change amount Xe is −1, the inclination data value Ks = −1 / 8. When the flat / noise area width data measurement value de = 0, the inclination data value Ks = Xe. Therefore, the inclination data value Ks is obtained as a value in the range from −1 to +1.

  It should be noted that in the pixels in the gradation flat region (Ar = 0) and the noise region (Ar = 1), the slope data value Ks at the immediately preceding non-noise gradation change position (Ar = 2) (that is, the non-starting point). (Which corresponds to the slope data value Ks obtained at the noise gradation change position). When the area signal Ar = 0 or 1, the slope data value Ks at the previous non-noise gradation change position is held as it is until the next non-noise gradation change position (Ar = 2).

  Therefore, the gradient data value Ks between the start point non-noise gradation change position (Ar = 2) and the end point non-noise gradation change position (Ar = 2) is set as the end point amplitude limit gradation change amount Xe. And the flat / noise area width data measured value de, and the slope of the gradation value used in the pixel (Ar = 0 or 1) that becomes the flat / noise area, the gradation flat area (Ar = 0) Together with the pixels, the pixels in the noise region (Ar = 1) are treated as pixels in a flat / noise region (a flat section when there is no noise component), and the slope is not calculated from the change due to the noise component. Therefore, the gradient of the gradation value in the pixel at the non-noise gradation change position (Ar = 2) can be obtained without erroneous detection due to noise components.

  FIGS. 30A and 30B are conceptual diagrams for explaining the inclination data value Ks calculated by the inclination detecting means 41 shown in FIG. FIG. 30A shows that when the end point amplitude limit gradation change amount Xe (height) is +1 and the tone value increases, FIG. 30B shows that the end point amplitude limit tone change amount Xe is −1. This shows a case where the gradation value decreases. 30A and 30B, the horizontal axis indicates the horizontal pixel position, and the vertical axis indicates the gradation value La (x) of the input image signal Da at each pixel position. As shown in FIGS. 30A and 30B, the non-noise gradation change position (Ar = 2) at the start point (pixel position x) and the non-noise gradation change position (Ar = 2) as the end point. If the flat / noise region width obtained from the number of pixels between the two is de, the pixel position at the end point position is x + de. From the flat / noise region width data measurement value de and the amplitude limit gradation change amount Xlm, FIG. ) And the slope Ks as shown in (b) is a pixel immediately before the non-noise gradation change position (Ar = 2) at the end point from the non-noise gradation change position (Ar = 2) at the start point (pixel position x). Is obtained as the slope data value Ks up to.

  As described above, the inclination detection unit 41 obtains the inclination data value Ks of the value based on the change of the gradation value at the non-noise gradation change position (Ar = 2), and the pixel (Ar = 0) that becomes the flat / noise area. Alternatively, in 1), the slope data value calculated at the previous non-noise gradation change position (Ar = 2) is held. Since the amplitude limit gradation change amount Xlm is limited within the range from −1 to +1, the detected slope data value Ks is a value within the range from −1 / de to + 1 / de, and thus − The value is in the range from 1 to +1.

  In FIG. 29, the area signal delay compensation means 67 receives the area signal Ar from the pixel area determination means 20. The area signal delay compensation unit 67 performs a non-noise gradation change position (Ar = 2) which is a start point in the flat / noise area width determination unit 80 with respect to the area signal Ar from a non-noise gradation change position (Ar = 2). The flat / noise area width data measured value de until Ar = 2) is obtained, and the area signal Ar is delayed for a time until the inclination data value Ks is obtained by the inclination detecting means 41, and the output ( The phase of the slope data value Ks) and the phase of the start point non-noise gradation change position are matched, and a delay compensated area signal Ard is output. The flat / noise area width determining means 80 and the inclination detecting means 41 obtain the inclination between the non-noise gradation change position and the next non-noise gradation change position, and the flat / noise area width and the end point between them are obtained. A delay is required to obtain the height of the gradation change. Therefore, the delay compensation time here is a time corresponding to the delay time until the flat / noise area width, the gradation change height, and the gradation change inclination are obtained.

  Next, in FIG. 29, the correction bit generation means 81 receives the inclination data value Ks from the inclination detection means 41 and the delay compensated area signal Ard from the area signal delay compensation means 67. The correction bit generation unit 81 obtains the gradation change amount dK2 for each adjacent pixel of the correction value (gradation value) to be added to the pixel value of the input pixel signal Da from the inclination data value Ks, and for each horizontal scanning period, In accordance with the area signal Ard (or area signal Ar), the gradation change amount dK2 based on the inclination data value Ks is sequentially added for each pixel to obtain a correction value, and correction bit data Ce is generated from the correction value, and pixel value calculation means Output to 94.

  Here, the inclination data value Ks, that is, the gradation change amount dK2 for each pixel is the gradation value La (for example, FIGS. 2A to 2C or FIGS. 3A to 3C) of the input image signal Da. The correction bit data Ce generated by the correction bit generation means 81 is added by a value including the decimal part of the gradation value of the input image signal Da. It is obtained as a bit string including the bit E and the sign bit of the value to be added. The correction bit data Ce is composed of an additional bit E and its sign bit, for example, like the correction bit data Cd described in FIG. 14 (Embodiment 1).

  FIG. 31 is a block diagram schematically showing an example of the configuration of the correction bit generation means 81 shown in FIG. As illustrated in FIG. 31, the correction bit generation unit 81 includes, for example, an addition unit 82, a switching unit 83, a correction value delay unit 84, and a bit string conversion unit 85. The correction bit generation means 81 obtains a correction value to be added to the input pixel signal Da, and outputs a bit string indicating the gradation value as correction bit data Ce.

  In the correction bit generation means 81 shown in FIG. 31, the inclination data value Ks from the inclination detection means 41 is input to the addition means 82, and a delay delayed by one pixel from the correction value delay means 84 described later. A correction value pCd is input. The adding means 82 obtains the gradation change amount dK2 for each adjacent pixel of the correction value from the inclination data value Ks, and adds the gradation change amount dK2 to the correction value in the immediately preceding pixel that is the delay correction value pCd. Thus, the correction value sCe is obtained for the pixel that becomes the gradation flat region or the noise region after the non-noise gradation change position (Ar = 2), and is output to the switching means 83. That is, the adding means 82 first obtains the inclination data value Ks as the gradation change amount dK2 for each adjacent pixel of the correction value (dK2 = Ks), and then the correction value (delay correction value) for the immediately preceding pixel. The gradation change amount dK2 is added to pCe) (pCe + dK2). The calculation of the adding means 82 is performed by adding or subtracting according to the sign of the gradation change amount dK2 (that is, the sign of the slope data value Ks) (when dK2> 0, the absolute value | dK2 | is added, When dK2 <0, | dK2 | is subtracted).

  The switching means 83 receives the area signal Ard (or area signal Ar) and the correction value sCe from the adding means 82. The switching unit 83 initializes the value for each horizontal scanning period (sets to a fixed start point correction value of 0), and based on the area signal Ar, the start point correction value of 0 and the correction value sCe of the addition unit 82 And a correction value aCe is generated and output as a gradation value to be added to the input image signal Da. More specifically, the switching means 83 selects the start point correction value 0 to set the correction value aCe = 0 at the time of initialization at the horizontal scanning period break, and in accordance with the area signal Ard, the non-noise gradation At the change position (Ard = 2), the start point correction value is 0 (correction value aCe = 0). On the other hand, in the pixel (Ard = 0 or 1) that is the gradation flat region or the noise region, the switching unit 83 uses the correction value sCe (that is, the gradation change amount dK2 with respect to the correction value pCe of the immediately preceding pixel). = A value obtained by adding Ks), and a correction value aCe is selected (aCe = pCe + Ks). The correction value aCe is output to the bit string conversion means 85 and sent to the correction value delay means 84, delayed by one pixel, and returned to the addition means 82 as a delay correction value pCe.

  The correction value delay means 84 delays the correction value aCe from the switching means 83 by one pixel and outputs a delay correction value pCe. Since the output from the switching unit 83 is delayed, the correction value for the immediately preceding pixel is output as the delay correction value pCe.

  Due to the configuration of the adding means 82, the switching means 83, and the correction value delay means 84 described above, the correction value aCe to be added to the input image signal Da starts at the non-noise gradation change position (Ard = 2). In a pixel (Ard = 0, 1) which has a point correction value of 0 and becomes a gradation flat region or a noise region, the correction value sCe by the adding means 82, that is, the immediately preceding position from the non-noise gradation change position (Ard = 2). The gradation change amount dK2 = Ks is added to the value in the pixel to obtain a value that sequentially changes with the magnitude of the absolute value of the gradation change amount dK2 (that is, the absolute value | Ks | of the gradient data value Ks). . The gradation change amount dK2 due to the inclination data value Ks is expressed as a decimal number whose absolute value in the range from −1 to +1 is 1 or less. Therefore, the correction value aCe obtained from the switching unit 83 is also 1 The gradation value is accurate to the following decimal.

  That is, with respect to the value N obtained by counting the pixels of the area signal Ar = 0 or 1 in order from the non-noise gradation change position (Ar = 2) that is the starting point, the correction value aCe for each pixel has a slope with the count value N The correction value aCe = N × Ks can be expressed by the data value Ks (gradation change amount dK2 = Ks), and the level of the gradation change amount dK2 due to the inclination data value Ks in order from the correction value aCe = 0 at the start point. A correction value aCe having a change in difference is obtained.

  Since the correction value aCe selected by the switching unit 83 is switched based on the area signal Ard (or Ar), the pixel that becomes the noise area (Ar = 1) as well as the gradation flat area (Ar = 0) has a flat period. A correction value aCe having a change in level difference of the gradation change amount dK2 due to the inclination data value Ks is obtained as a pixel, and thus a correction value aCe that changes sequentially without being affected by the noise component.

  Note that the gradient data value Ks is limited within a range from −1 to +1, and has a gradation change amount in which a gradation change such as an edge (outline) portion exceeds the range from −1 to +1. Even in the pixel, the added correction value is in the range from −1 to +1, and the difference is held. Therefore, it can be obtained as a correction value that preserves the contour information.

  The correction value aCe from the switching unit 83 is input to the bit string conversion unit 85 in the correction bit generation unit 81. The bit string conversion means 85 converts the correction value aCe into a bit string value having a predetermined number of bits, obtains the correction bit data Ce including the additional bit E indicating the value to be added and the sign bit of the value, and correction bit generation means 81 to the pixel value calculation means 63. The bit string conversion means 85 can be configured in the same manner as the bit string conversion means 55 in FIG. 13 (Embodiment 1) only with different inputs, but the absolute value excluding the sign of the correction value aCe is a value including a decimal number of 1 or less. Thus, the correction value (additional bit E) in the correction bit data Ce indicates a decimal part with respect to the bit string of the gradation value of the input image signal Da. Similarly to the case of FIG. 13, the bit string converting means 85 can be omitted because the value such as the correction value aCe is usually obtained as a value of a predetermined number of bits in hardware. Thus, it is possible to omit the case where it is not necessary to treat it as a bit string.

  FIG. 32 is a block diagram schematically showing another example of the configuration of the correction bit generation means 81b shown in FIG. As shown in FIG. 32, the correction bit generation unit 81b counts pixels 86 in order from the non-noise gradation change position (Ar = 2) serving as a start point with the region signal Ard (or Ar) as a reference. And a multiplication means 87 for multiplying the count value N and the inclination data value Ks to obtain the correction value aCe = N × Ks, and a bit string conversion means 85. The correction bit generation unit 81b obtains a correction value aCe having a change in level difference due to the inclination data value Ks in order from the correction value aCe at the starting point = 0.

  Next, in the delay compensation means 93 in the gradation conversion means 4 shown in FIG. 29, the delay until the correction bit data Ce output from the correction bit generation means 81 is obtained matches the pixel position of the input image signal Da. Delay compensation of the input image signal Da is performed, and a delay compensated image signal Dd is output.

  The pixel value calculation unit 94 in the gradation conversion unit 4 receives the delay compensated image signal Dd from the delay compensation unit 93 and the correction bit data Ce from the correction bit generation unit 81. The pixel value calculation means 94 calculates a new output image signal Db by adding the correction bit data Ce from the correction bit generation means 81 to the delay compensated image signal Dd with the position of the integer part aligned. ,Output. Since correction bit data Ce including a decimal part is added to the gradation value of the output image signal Db by calculation, the number of bits of the output image signal Db is larger than the number of bits of the image signal Dd, that is, expanded. The output image signal Db with the expanded number of bits becomes the output of the gradation converting means 4.

  In the calculation in the pixel value calculation means 94, addition or subtraction processing is performed according to the positive or negative sign of the correction bit data Ce (that is, when Ce> 0, the absolute value | Ce | is added, and Ce < When 0, | Ce | is subtracted). Further, when the value after the calculation is a negative value, the output image signal Db is set to the value 0 (Db = 0), and when the value exceeds the set bit number (maximum value) of the output image signal Db, The value of the output image signal Db is limited by setting the value of the output image signal Db as a predetermined maximum value.

  The output image signal Db after being processed by addition or subtraction in the pixel value calculation means 94 is an image signal with a precision having a decimal part with respect to the image signal Dd, and the number of bits of the output image signal Db is the input image signal. Since it is larger than Da, not only the number of gradations that can be expressed is increased, but also the number of bits in the decimal part of the correction bit data Ce including the decimal part is included, the resolution of the gradation value is increased, and the smoother gradation It will be possible to express changes in values. Further, even when the change in the gradation value is caused by the edge (contour) of the image, the difference in the edge portion is retained, so that the smoothness of the image can be increased while maintaining the contour information.

  Further, the correction bit data Ce is based on the area signal Ar from the pixel area determination unit 20 and is not affected by noise components, and the gradation change amount based on the gradient data value Ks in order from the non-noise gradation change position (Ar = 2). The correction value aCe that changes with dK2 is obtained, and the pixel value calculation means 94 adds or subtracts the value based on the correction bit data Ce to the gradation value of the image signal Dd (or the input image signal Da). Therefore, compared to the gradation value conversion processing by linear interpolation processing, the image details due to slight changes in gradation values such as noise remain, and sharpness is maintained while maintaining the sharpness of pixels whose gradation values change slowly. The smoothness of the area can be increased, and blurring of the image can be reduced.

  The relationship among the input image signal Da, the correction bit data Ce, and the output image signal Db in the fourth embodiment is the same as in the case of FIG. 14 (first embodiment).

<< 4-2 >> Operation.
FIG. 33 is a flowchart showing the operation (No. 1) of the image processing apparatus according to the fourth embodiment, and mainly shows the operation for calculating the inclination data value by the gradation converting means 4. FIG. 34 is a flowchart showing the operation (part 2) of the image processing apparatus according to the fourth embodiment, and shows the operation of the correction bit data generation and gradation conversion processing by the gradation converting means 4. 33 and 34 show processing for each pixel in each scanning line (line). Note that, based on the gradation change amount df in the pixel area determination unit 20, each pixel is classified into one of a pixel belonging to a flat gradation area, a pixel belonging to a noise area, and a pixel at a non-noise gradation change position. The operation is the same as in FIG. 17 (Embodiment 1).

  FIGS. 35A to 35K are diagrams for explaining an example of the operation of the image processing apparatus according to the fourth embodiment (when the gradation value rises very slowly and includes a noise component). FIGS. 36A to 36K are diagrams for explaining another example of the operation of the image processing apparatus according to the fourth embodiment (when the gradation value decreases very slowly and includes a noise component). FIG. FIGS. 35A to 35K show the operation of the gradation converting means 4 for a part of the input image signal Da including the noise component similar to that shown in FIGS. 18A to 18K (Embodiment 1). 36 (a) to 36 (k) show the operation in the gradation converting means 4 when the same image signals as those in FIGS. 19 (a) to 19 (k) (Embodiment 1) are input. ing.

  Hereinafter, the operation of the gradation converting means 4 will be described with reference to FIGS. 33, 34, 35 (a) to (k), and FIGS. 36 (a) to (k). First, in FIG. 33, the operations in steps S101 to S103, S105, that is, the input image signal Da is input, the gradation change amount df is obtained, the amplitude limited gradation change amount Xlm is obtained, and the pixel region determination result is obtained. The operation for generating the region signal Ar shown is the same as the operation shown in FIG. 15 (Embodiment 1). 35A to 35E and FIGS. 36A to 36E are the same as FIGS. 18A to 18E and FIGS. 19A to 19E, respectively.

  Next, the flat / noise area width determination means 80 and the inclination detection means 41 obtain the flat / noise area width data measurement value de between the non-noise gradation change positions, and detect the inclination of the gradation value. This is performed at the non-noise gradation change position (Ar = 2) (steps S200 to S205).

  At the non-noise gradation change position (Ar = 2) (step S200), the flat / noise area width determining means 80 starts from this pixel position (step S201), and then the non-noise gradation change position (Ar = The number of pixels up to the end point pixel position 2) is counted and a count value CN is output (step S202). Then, a value (CN + 1) obtained by adding 1 to the obtained count value CN is set as a flat / noise area width data measurement value de (step S203). The flat / noise area width data measurement value de is held from the non-noise gradation change position at the start point to the pixel immediately before the non-noise gradation change position as the end point. That is, in the pixel in the flat gradation region (Ar = 0) and the pixel in the noise region (Ar = 1), the value at the non-noise gradation change position at the immediately preceding start point is maintained, and the gradient of the gradation value between them The width (distance) at (step S206).

  In FIG. 35 (f), the count value CN is reset to 0 at the pixel position x = 8, 16, 24, 32 which is the non-noise gradation change position (Ar = 2), and the next non-noise gradation change position. The number of pixels up to (Ar = 2) is counted, and the continuous number of pixels that become the gradation flat region (Ar = 0) and the noise region (Ar = 1) is obtained. In FIG. 35 (g), the flat / noise area width data measurement value de obtained from the value obtained by adding 1 to the count value CN is the non-noise gradation change that is the end point from the non-noise gradation change position that is the start point. It is held until the position immediately before the position. 36 (f) and (g) are the same as those in FIGS. 35 (f) and (g).

  The inclination detecting means 41 obtains the amplitude limit gradation change amount Xlm at the end point non-noise gradation change position (Ar = 2) as the end point amplitude limit gradation change amount Xe as the end point amplitude limit gradation change amount Xe (step S204). ), The gradation value from the non-noise gradation change position as the start point to the non-noise gradation change position as the end point from the end point amplitude limit gradation change amount Xe and the flat / noise area width data measurement value de. The inclination is calculated as an inclination data value Ks = Xe / de (step S205). In the gradation flat region (Ar = 0) and noise region (Ar = 1), the value at the previous non-noise gradation change position (Ar = 2) is held (step S206). The amplitude limit gradation change amount Xlm, that is, the amplitude limit gradation change amount Xe at the end point is input as a value in which the gradation change amount df is limited within a range from −1 to +1. The slope data value Ks of the tone value is a value in the range from −1 to +1.

  According to FIG. 35 (h), at the pixel position x = 8 which is the non-noise gradation change position (Ar = 2), the flat / noise area width data measurement value de = 8, the end point amplitude limit gradation change amount Xe. = + 1, the slope data value Ks = + 1/8, and similarly, the slope data value Ks is obtained at the pixel position x = 16, 24, 32. Also, according to FIG. 36 (h), at the pixel position x = 8 which is the non-noise gradation change position (Ar = 2), the flat / noise area width data measurement value de = 8, the end point amplitude limit gradation change. With the amount Xe = −1, the inclination data value Ks = −1 / 8, and similarly, the inclination data value Ks is obtained at the pixel position x = 16, 24, 32. In both FIG. 35 (h) and FIG. 36 (h), the values at the pixel positions x = 8, 16, 24, and 32 are the same in the gradation flat region (Ar = 0) and the noise region (Ar = 1). Is retained.

  With the above operation shown in FIG. 33, the flat / noise area width data measurement value de and the inclination data value Ks in the gradient of the gradation value are obtained with reference to the area signal Ar, and the non-noise gradation is obtained without erroneous detection due to the noise component. The gradient of the gradation value in the pixel at the change position (Ar = 2) can be obtained.

  Next, in FIG. 34, after obtaining the gradient data value Ks of the gradation value according to the area signal Ar, the correction bit generation means 81 initializes the correction value aCe at the boundary of the horizontal scanning period, and The processing for generating the correction value aCe is switched according to the area signal Ar indicating (step S210).

  At the non-noise gradation change position (Ar = 2) (step S210), the correction value aCe = 0 is set so that the correction value is 0 at the start point (step S211).

  On the other hand, in a pixel that becomes a gradation flat region (Ard = 0) or a noise region (Ard = 1), the gradient data value Ks is set as the gradation change amount dK2 for each adjacent pixel of the correction value (dK2 = Ks). ) The gradation change amount dK2 is added (pCe + dK2) to the correction value (delay correction value pCe) in the immediately preceding pixel to obtain the correction value sCe in the pixel in the gradation flat region (step S212). Then, the correction value sCe is determined as the correction value aCe (step S213). As a result, the correction value sCe for the pixels in the flat gradation region (Ar = 0) and the noise region (Ar = 1) is the magnitude of the absolute value of the gradation change amount dK2 in order from the change position (that is, the slope data value Ks). The absolute value of | Ks |) changes. Since the slope data value Ks is a value in the range from −1 to +1, the correction value sCe is also expressed as a decimal number whose absolute value in the range from −1 to +1 is 1 or less.

  The correction value aCe obtained as described above is converted into a bit string value having a predetermined number of bits, and is determined as correction bit data Ce including an additional bit E indicating a value to be added and a sign bit of the value (step S214). ). The correction value aCe is delayed by one pixel and sent to the next pixel process as the correction value pCe for the immediately preceding pixel (step S215).

  Through steps S210 to S214, the correction value aCe (that is, the correction bit data Ce) to be added to the input image signal Da becomes the correction value aCe = 0 at the non-noise gradation change position (Ar = 2), and the gradation In the flat region (Ar = 0) and the noise region (Ar = 1), values are obtained as values that sequentially change with a level difference due to the slope data value Ks in order from the value 0 at the non-noise gradation change position (Ar = 2). .

  Since the correction value aCe to be obtained is switched based on the area signal Ar, the pixels that become the noise area (Ar = 1) together with the gradation flat area (Ar = 0) are also defined as the pixels in the flat period (gradation flat area ( The processing is the same as that of Ar = 0). A correction value aCe having a change in level difference due to the inclination data value Ks is obtained, and a correction value aCe that changes sequentially without being affected by the noise component is obtained.

  According to FIG. 35 (i), the correction value aCe = 0 at the pixel position x = 16 (Ar = 2), and the gradation flat region (Ar = 0) and noise region (Ar = 1) thereafter. In the order of pixels, the correction value aCe (that is, the correction bit data Ce) having a change of the level difference + 1/8 due to the inclination data value Ks, such as 1/8, 2/8,. can get. Further, according to FIG. 36 (i), the correction value aCe = 0 at the pixel position x = 16 (Ar = 2), and the subsequent gray level flat area (Ar = 0) and noise area (Ar = In the pixel 1), the correction value aCe (that is, the correction bit data Ce) having a change of the level difference of 1/8 by the inclination data value Ks, such as −1/8, −2/8,. ).

  The pixel value calculation means 94 adds the correction bit data Ce with the position of the integer part matched to the image signal Dd (or Da) that has been delay compensated so as to match the pixel position of the correction bit data Ce and the input image signal Da. Thus, a new output image signal Db is calculated (step S216). When the value after the calculation is a negative value, the output image signal Db is set to zero (Db = 0), and when the value exceeds the set bit number (maximum value) of the output image signal Db, Is the maximum value, and the value of the output image signal Db is limited.

  Since the correction bit data Ce includes a decimal part with respect to the gradation value of the image signal Dd (or the input image signal Da), the output image signal Db after the arithmetic processing has a decimal part with respect to the image signal Dd. It becomes an image signal with a certain accuracy. Since the output image signal Db has the number of bits in the decimal part of the correction bit data Ce including the decimal part, the resolution of the gradation value is increased, and a smoother change in the gradation value can be expressed. The correction bit data Ce is obtained as a correction value aCe that changes with the gradient data value Ks without being affected by the noise component on the basis of the area signal Ar, and the value of the correction bit data Ce is added to the gradation value or Since the subtraction processing is performed, it is possible to increase the smoothness of the pixel region in which the gradation value gradually changes while leaving the image detail and sharpness due to a slight change in the gradation value such as noise.

  According to FIG. 35 (j), the gradation value Lb of the output image signal Db after the addition of the correction bit data Ce is made non-noise by adding the correction bit data Ce to the gradation value La of the image signal Dd. At the pixel position x = 16, which is the gradation change position (Ar = 2), the gradation value Lb = 24/8, and in subsequent gradation flat areas and noise areas, the pixel position increases by 1/8 in order. In the pixel position x = 19 where there is a noise component and there is a noise component, the gradation value is obtained as a converted value while leaving the amplitude due to the noise component. FIG. 36 (j) is the same as FIG. 35 (j). 35 (k) and 36 (k) show the output image signal Db when the additional bit E in the correction bit data Ce is 4 bits, from the non-noise gradation change position (Ar = 2). A 12-bit output image signal Db obtained by changing the gradation change more smoothly is obtained as a gradation value having a change in value corresponding to the inclination data value Ks in the order of the pixel position.

  In the above description, the decimal part of the additional bit E in the correction bit data Ce is 4 bits. However, it may be other than 4 bits. For example, the number of bits in the decimal part is increased to increase the number of bits. By expressing the correction bit data Ce (that is, the correction value aCe), it is possible to obtain an image signal with a more accurate and smoother gradation value change.

  FIG. 37 is a flowchart showing another example of the operation of the image processing apparatus according to the fourth embodiment (an example instead of FIG. 34). Generation of correction bit data and gradation by the gradation converting means 4 shown in FIG. The operation of the conversion process is shown. When generating the correction bit data, the correction value is calculated from the count value N and the inclination data value Ks obtained by counting the pixels in order from the non-noise gradation change position (Ar = 2) as the start point with the area signal Ar as a reference. It can also be determined as aCe = N × Ks. In this case, if steps S220 to S222 in FIG. 37 are performed instead of steps S211 to S213 and step S215 in FIG. 34, the correction bit data Ce can be generated as in the case of FIG. A new added output image signal Db can be obtained. That is, at the non-noise gradation change position (Ar = 2) (step S210), the pixel count value N = 0 (step S220), and the pixel (Ard) which becomes the gradation flat area (Ard = 0) or the noise area = 1), the number of pixels is counted and set to N + 1 (step S221). The correction value aCe is multiplied by the count value N and the inclination data value Ks to obtain a correction value aCe = N × Ks (step S222). As a result, as in the case of FIG. 34, the correction value aCe having a change in level difference due to the inclination data value Ks can be obtained in order from the correction value aCe at the starting point = 0.

<< 4-3 >> Effect.
As described above, in the image processing apparatus or the image processing method according to the fourth embodiment, the non-noise gradation change position (Ar = 2) is determined based on the region signal Ar obtained from the gradation change amount df. A flat / noise area width data measurement value de is obtained, and a gradient data value Ks of the gradation value is obtained from the amplitude limit gradation change amount Xlm and the flat / noise area width data measurement value de. Then, the correction bit generation unit 81 generates correction bit data Ce from the inclination data value Ks, and the pixel value calculation unit 94 adds the value of the correction bit data by aligning the position of the integer part with the input image signal Da. An image signal with an expanded number is obtained. Therefore, according to the image processing apparatus or the image processing method of the fourth embodiment, the flat / noise area width indicating the width between the non-noise gradation change position and the next non-noise gradation change position is accurately determined. Since it can be detected and can be distinguished from the gradation change due to the noise component, the inclination data value can be calculated without being influenced by the noise component. Further, according to the image processing apparatus or the image processing method of the fourth embodiment, the correction value that sequentially changes with the gradation value based on the inclination data value Ks can be obtained without significantly increasing the circuit scale, and the gradation value gradually increases. Thus, it is possible to obtain appropriate correction bit data Ce in a pixel region where the change occurs or in a non-noise gradation change position, to convert the gradation change of the image signal more smoothly, and to expand the number of bits. In addition, since the obtained correction bit data Ce is added to the gradation value of the input image signal Da to extend the number of gradations, the number of gradations can be processed in real time. Changes can be converted more smoothly.

  As described above, according to the image processing apparatus or the image processing method of the fourth embodiment, even when the input image signal Da includes a noise component, the gradation is gradually increased without significantly increasing the circuit scale in real time. It is possible to obtain the correction bit data Ce obtained by more appropriately detecting the gradation change in the image region whose value changes without erroneous detection due to the noise component. For this reason, the number of gradations of the image signal is expanded by the correction bit data, and the image details are not lost or the sharpness is not lost in the gradation-converted output image signal Db. Alternatively, the pseudo contour is reduced, the gradation change is smooth, and an image signal with an extended number of gradations can be obtained.

<< 4-4 >> Modifications.
In the above description, the case where the region signal Ar is obtained by the pixel region determination unit similar to the pixel region determination unit 20 (FIG. 5) in the first embodiment has been described. The area determination means 20b (FIG. 22) or the pixel area determination means 70 and 70b (FIGS. 25 and 28) in the third embodiment may be used. In this case, the same effect can be obtained.

  In the above description, three types of region signals Ar = 0, 1, and 2 are generated as region determination results in the pixel region determination unit 20, but the flat / noise region width determination unit 80 in FIG. The detection means 41 and the correction bit generation means 81 (or 81b) also perform processing in two ways: a non-noise gradation change position (Ar = 2), a gradation flat area, and a noise area (Ar = 0, 1). Switching. For this reason, the region determination result may be configured as a region signal Ar1 that is a binary value of a non-noise gradation change position, a gradation flat region, or a noise region, and the same effect can be obtained.

  In the above description, the case where the gradation change amount detecting means 10 limits the gradation change amount df due to the difference in gradation value between adjacent pixels in the horizontal direction to the predetermined range −1 and +1 has been described. The present invention is not limited to such a form. For example, when a change in a range from the maximum value +2 to the minimum value −2 of the gradation value difference between pixels is obtained, a predetermined maximum value (positive Value) and the minimum value (negative value), the gradation change amount df is limited, and the gradation value changes within a gradation difference of a value larger than one gradation of the image signal. It can also be configured to extend the logarithm. The range for limiting the value of the gradation change amount df is a value for determining whether the image area is a smooth change by gradation conversion, whether it is retained as it is, or whether it is corrected as a gradation change that becomes a gradation jump. An appropriate value may be set from the characteristics of the input image signal, the level at which the gradation jump occurs, and the like. The configuration and operation in this case are the same as the configuration and operation shown in FIG. 29 except for the range of the difference in gradation value between pixels for which a change in gradation value is to be obtained. Here, in the case of obtaining a change in the range from the maximum value +2 to the minimum value −2 of the gradation value difference between the pixels, the restriction range in the gradation change amount limiting unit 12 in the gradation change amount detection unit 10. That is, the value of the amplitude limit gradation change amount Xlm is a value in the range from +2 to −2, and the gradient of the gradation value is also a value in the range from −2 to +2. In the range set by the maximum value (positive value) and the minimum value (negative value), such as when the change in the range from the maximum value +2 to the minimum value -2 of the gradation value difference between pixels is obtained, Even in the case where the tone change is detected, the tone converting means 4 can be applied, and the same configuration can be obtained and the same effect can be obtained.

  Further, the number of gradations m of the input image signal Da is not limited to 8 bits, and may be, for example, a 10-bit image signal. In the pixel value calculation unit 94, an integer of additional bits E in the correction bit data Cd. The same effect can be achieved by calculating the value in which the position of the part matches the position of the integer part of the image signal Dd.

  In the above description, the gradation change amount limiting means 12 in the gradation change amount detection means 10 limits the value of the gradation change amount df so that it falls within the range from −1 to +1. As described above, when the absolute value of the gradation change amount df is larger than a predetermined value (for example, | df |> 1), the gradation change amount df is a change in the gradation value due to an edge (outline or the like) of the image. In this way, when the gradation change is determined to be due to an edge, the gradation value is not added as the correction value aCe = 0, and the contour information of the image is maintained. You can also That is, in the gradation change amount limiting means 12 in the gradation change amount detection means 10, the amplitude limited gradation change whose value is limited so that the value of the gradation change amount df is in the range from −1 to +1. When the value of the gradation change amount df is out of the range from −1 to +1 when the amount Xlm is obtained, the value is set to 0 (that is, the amplitude limit gradation change amount Xlm = 0). ) Even in the configuration, in the region having the non-noise gradation change position of the contour as the end point pixel, the inclination data value Ks = 0 and the correction value aCe is also 0, so that no gradation value is added, Since gradation conversion is not performed, changes in gradation values in an image region where gradation values change gradually other than changes that become outlines can be made smoother while maintaining the sharpness (contour information) of the image.

  In the above description, the gradation converting means 4 has been described as a configuration that detects and processes the amount of gradation change in the horizontal direction in the change of the gradation value in the horizontal direction of the input image signal Da. The gradation change amount in the vertical direction (screen display scanning line direction) of Da is detected, correction bit data Ce is generated and added to the image signal, and the gradation change in the vertical direction may be smoothed. There is an effect. In this case, the tone change amount detection means 10 obtains the tone change amount df from the difference between the pixels on the upper scanning line positioned vertically above and below, and the pixel region determination means 20 determines the pixel region. Then, the processing in the flat / noise area width determining means 80, the inclination detecting means 41, and the correction bit generating means 81 is processed in the line (one horizontal scanning period) direction, and the gradation values in the pixels positioned vertically above and below What is necessary is just to comprise so that the change of may be detected.

  Further, in the above description, it has been described that each configuration of the gradation converting means 4 in the image processing apparatus is a hardware configuration, but the present invention is configured to be realized by software processing in program control. Also good.

  In the fourth embodiment, points other than the above are the same as in any of the first to third embodiments.

<< 5 >> Embodiment 5
The image processing apparatus according to any of the first to fourth embodiments detects a change in gradation value in one of the horizontal direction and the vertical direction (constant direction) of the image signal and detects the gradient of the gradation value (for example, the gradient). Data value Ksl) is obtained, and gradation conversion processing for expanding the number of bits of the input image signal Da using the obtained gradient is performed, thereby generating an output image signal Db having a higher gradation than the input image signal Da. It is configured as follows. In contrast, the image processing apparatus according to the fifth embodiment of the present invention detects a change in gradation value in each of the horizontal direction and the vertical direction, and obtains a gradient of the gradation value in each of the horizontal direction and the vertical direction. By performing gradation conversion processing that expands the number of bits of the input image signal Da using the obtained gradient, an output image signal Dc having a multi-gradation than the input image signal Da is generated, and the two-dimensional direction is generated. The gradation value is smoothly changed.

  FIG. 38 is a block diagram schematically showing an example of the configuration of an image processing apparatus according to the fifth embodiment of the present invention (that is, an apparatus capable of performing the image processing method according to the fifth embodiment). The image processing apparatus according to the fifth embodiment detects a gradation change amount in each of the horizontal direction and the vertical direction of the image signal, and uses the detected gradation change amount to generate pixel regions in the horizontal direction and the vertical direction. Is classified into one of the pixels belonging to the flat gradation region, the pixel belonging to the noise region, or the pixel at the non-noise gradation change position, and the gradient of the gradation value is determined based on the determination result of the pixel region. And the gradation of the input image signal Da is expanded using the obtained gradient of the gradation value, and gradation conversion processing is performed to smooth the change of the gradation value in the two-dimensional direction. .

  As shown in FIG. 38, the image processing apparatus according to the fifth embodiment is adapted to the horizontal and vertical directions for the m-bit digital input image signal Da from the respective gradation change amounts in the horizontal and vertical directions. Each pixel is determined to be classified as either a pixel belonging to a flat gradation region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position, and a gradation value is determined based on the determination result of this pixel region. Is detected, and a gradation conversion process for extending the number of bits is performed by adding a value corresponding to the gradient of the gradation value, and output as an n-bit output image signal Dc, and the gradation value in the two-dimensional direction Is provided with gradation conversion means 5 for performing gradation conversion processing for smoothing the change of the above. The gradation converting means 5 detects a change in the gradation value in the horizontal direction and adds a correction value according to the gradient of the gradation value to extend the number of bits, and Vertical direction gradation conversion means 111 is provided that detects a change in gradation value in the vertical direction and performs gradation conversion processing using a correction value corresponding to the gradient of the gradation value.

  In the following description, the gradation converting means 5 determines the difference of one gradation value (+1) from the difference of gradation values between adjacent pixels in the horizontal direction and the vertical direction of the input image signal Da. And -1) are detected to obtain the gradient of the gradation value, and gradation conversion processing is performed to expand the number of gradations using the obtained gradient, and the 8-bit input image signal Da is converted to 12 A case where the number of gradations is expanded and output to the bit output image signal Dc will be described.

  The gradation converting means 5 detects a change in gradation value in each of the horizontal direction and the vertical direction with respect to the 8-bit input image signal Da, and obtains the gradient of the gradation value, for example, for 4 bits. Correction bit data that is the value of the correction value is obtained, arithmetic processing is performed on the gradation value of the image signal, and gradation conversion is performed by extending the number of bits by adding bits to produce a 12-bit output image Output as signal Dc.

  The gradation converting means 5 will be described in more detail. First, the input image signal Da input to the gradation conversion unit 5 is input to the horizontal direction gradation processing unit 110 in the gradation conversion unit 5. The horizontal direction gradation processing unit 110 is configured in the same manner as the gradation conversion units 1 to 4 in the first to fourth embodiments, and is based on the difference in gradation value between pixels in the horizontal direction of the input image signal Da. A tone change amount is obtained, and each pixel is classified into a pixel belonging to a flat gradation region, a pixel belonging to a noise region, or a pixel at a non-noise gradation change position based on the obtained gradation change amount value. Make a decision. Then, based on the determination result of the pixel area, the flat / noise area width is obtained, the gradient of the gradation value is obtained from the flat / noise area width and the gradation change amount, and the gradient of the obtained gradation value is determined. By adding the corrected bit data to the input image signal, the number of bits is expanded, and a 12-bit output image signal Db with the expanded number of bits is generated. The gradation converting means 5 determines the pixel area for the gradation change in the horizontal direction of each pixel and obtains the flatness / noise area width and inclination of the gradation value based on the determination result, thereby preventing erroneous detection due to noise components. A change in the gradation value can be detected, and the gradation value in the horizontal direction can be expanded to a sufficient number of gradations to smooth the gradation change. As described above, the horizontal direction gradation processing unit 110 is the same as any one of the gradation conversion units 1 to 4 of the first to fourth embodiments.

  The 12-bit image signal Db output from the horizontal direction gradation processing unit 110 is input to the vertical direction gradation processing unit 111 in the gradation conversion unit 5. The vertical gradation processing means 111 is an upper 8 bits of the 12-bit image signal Db output from the horizontal direction gradation processing means 110 excluding the lower 4 bits corresponding to the decimal part of the input image signal Da. From the signal, the gradation change amount by the difference in gradation value between the pixels in the vertical direction is obtained, and the pixels belonging to the flat gradation region and the pixels belonging to the noise region are determined based on the obtained gradation change amount value. , Classify the pixel into one of the non-noise gradation change positions, and based on the determination result of this pixel area, the flat / noise area width in gradation change, and the flat / noise area width and gradation change amount Then, the gradient of the gradation value is obtained, and a correction value corresponding to the gradient is obtained. Then, gradation conversion is performed by adding correction bit data equivalent to that in the horizontal gradation processing means 110 to the image signal Db, and the result is output as a 12-bit image signal Dc. As a result, even in an image region where the gradation value gradually changes not only in the horizontal direction but also in the vertical direction, the change in the gradation value is detected, and the gradation change is smooth and the high-quality 12-bit image signal Dc. Is output as

  The vertical gradation processing unit 111 is configured in the vertical direction (up and down direction of the screen) in the gradation change amount detection unit 10 in any one of the gradation conversion units 1 to 4 of the first to fourth embodiments. A gradation change amount df is obtained by a difference in gradation values between pixels (difference between a gradation value of a target pixel and a gradation value of an adjacent pixel on a scanning line that is vertically upward by one scanning line) The area determination means 20 (or 20b, 70, 70b) determines the pixel area in the vertical direction, and the flat / noise area width determination means 30 (or 80), the inclination detection means 40 (or 41), and the correction bit generation means 50. The processing within (or 81) may be configured to detect a change in gradation value in a pixel located vertically above and below as processing in the line (one horizontal scanning period) direction.

  Further, when detecting a change in gradation value in the vertical direction and determining a gradient of the gradation value by determining a pixel area in the vertical direction, in order to detect a change in one gradation of the input image signal Da. Of the 12-bit image signal Db from the input horizontal direction gradation processing means 110, a value represented by the upper 8 bits excluding the lower 4 bits corresponding to the decimal part may be used.

  For this reason, in the description of the vertical direction gradation processing unit 111, as in any of the first to fourth embodiments, the pixel region is determined, the flatness / noise region width in the gradation change is obtained, and the gradient of the gradation value is determined. And the processing for obtaining the correction bit data using this inclination may be performed.

  From the above, the gradation converting means 5 shown in FIG. 38 determines the pixel area of each pixel based on the gradation change amount df of the gradation change in the horizontal direction in the horizontal gradation processing means 110. Based on the result, the gradation is converted by obtaining the flat / noise area width and inclination in the gradation change in the horizontal direction, and the gradation of the gradation change in the vertical direction is similarly applied to the vertical direction gradation processing unit 111. The pixel area is determined based on the amount of change, and based on this determination result, the gradation value is converted by obtaining the flat / noise area width and gradient in the gradation change in the vertical direction. For this reason, even when a noise component is included, the flatness / noise area width of the gradation change can be obtained in each of the two-dimensional directions, and the inclination can be obtained without erroneous detection due to the noise component. Appropriate correction bit data is obtained when it is determined that the change in the area of the changing pixel or the gradation value is due to the edge, and the 12-bit of which the gradation change in the vertical direction as well as the horizontal direction is more smoothly converted. An image signal Dc is obtained. In addition, since the obtained correction bit data is added or subtracted from the gradation value of the input image signal Da to extend the number of gradations, it can be processed in real time, resulting in loss of image details and sharpness. A 12-bit image signal Dc in which the gradation change is converted more smoothly without being damaged is obtained.

  In the description of the gradation conversion means 5 in FIG. 38, the gradation change amount when detecting the gradation change of the input image signal Da is the difference of the least significant 1 bit, that is, the gradation value between pixels. Although the difference is +1 and −1, the present invention is not limited to such a form. In each of the horizontal direction gradation processing unit 110 and the vertical direction gradation processing unit 111 in the gradation conversion unit 5, Based on the determined maximum value (positive value) and minimum value (negative value), the gradation change amount df is limited, and the gradation is within a gradation difference of a value larger than one gradation of the image signal. When the value changes, the number of gradations may be extended.

  In addition, the image signal often has high horizontal resolution. For this reason, the gradation converting means 3 in FIG. 38 is also configured to perform the vertical direction processing after the horizontal direction processing is first performed by the horizontal direction gradation processing means 110, so that a gradation change is detected. For example, a difference between two gradations (a gradation value between pixels) in the horizontal direction is set as a different value between the horizontal direction and the vertical direction. Difference +2 and -2), a difference of one gradation in the vertical direction, that is, a difference of gradation values +1 and -1 between pixels, and a difference of gradation detected in the horizontal direction. The weighting may be increased with respect to the vertical direction to expand the number of gradations.

  As described above, in the image processing apparatus according to the fifth embodiment, the horizontal gradation processing unit 110 determines the pixel area of each pixel based on the gradation change amount df of the horizontal gradation change. Based on this determination result, the gradation value is converted by obtaining the flat / noise area width and inclination in the gradation change in the horizontal direction, and the vertical direction gradation processing means 111 similarly applies the gradation in the vertical direction. The pixel area of each pixel is determined based on the amount of change in gradation, and the gradation value is converted by obtaining the flat / noise area width and inclination in the gradation change in the vertical direction based on the determination result. Therefore, even when a noise component is included in the input image signal Da, it is possible to obtain the flatness / noise area width and inclination of the gradation change in each of the two-dimensional directions without erroneous detection due to the noise component, and to increase the circuit scale. Correction values that change sequentially according to the gradation value obtained by the obtained gradient can be obtained without increasing significantly, and it is determined that the pixel area in which the gradation value changes gradually in the two-dimensional direction or the change in gradation value is due to the edge. In this case, appropriate correction bit data is obtained, and the number of gradations of the image signal can be expanded so that the gradation change in the vertical direction as well as the horizontal direction becomes smoother. Further, since the correction bit data obtained in each direction is added to the gradation value of the input image signal Da to extend the number of gradations, the gradation conversion process can be performed in real time, and noise can be processed. It is possible to convert a very gradual gradation change into an output image signal Dc that can be displayed more smoothly while leaving a slight change due to the above.

  Therefore, according to the image processing apparatus or the image processing method of the fifth embodiment, even when a noise component is included in the input image signal Da, the gradation value is gradually increased without significantly increasing the circuit scale in real time. Correction bit data Cd obtained by more appropriately detecting a gradation change can be obtained for a changing image region without erroneous detection due to a noise component. For this reason, the number of gradations of the image signal is expanded by the correction bit data, and in the output image signal Db subjected to gradation conversion, the image details are not lost or the sharpness is not lost, and the two-dimensional direction Quantization noise or pseudo contour is reduced, and a gradation change is smooth, and an image signal with an extended number of gradations can be obtained.

  In the image processing apparatus according to the fifth embodiment, the number m of gradations of the input image signal Da is not limited to 8 bits, and may be, for example, a 10-bit image signal. If the means 110 and the vertical direction gradation processing means 111 are configured to calculate a value in which the position of the integer part of the additional bit E in the correction bit data matches the position of the integer part of the input image signal Da. The same effect can be produced.

  Further, in the description of the fifth embodiment, it is described that the configuration of the gradation converting means 3 is a hardware configuration. However, the present invention is configured to be realized by software processing in program control. Also good.

  In the fifth embodiment, points other than those described above are the same as in any of the first to fourth embodiments.

<< 6 >> Embodiment 6
In the first to fifth embodiments, the image processing apparatus and the image processing method to which the present invention is applied have been described, but the present invention can also be applied to an image display apparatus. FIG. 39 is a block diagram schematically showing an example of the configuration of the image display device 6 according to the sixth embodiment of the present invention. As shown in FIG. 39, the image display device 6 of the sixth embodiment includes an image signal input terminal 501, a reception processing unit 502, a gradation conversion unit 500, and a display unit 504. The gradation conversion unit 500 is configured by, for example, the gradation conversion units 1 to 5 in the image processing apparatuses according to the first to fifth embodiments. The gradation conversion unit 500 is configured to extend the number of bits of the input image signal and output the gradation-converted multi-gradation image signal to the display unit 504.

  The input terminal 501 receives a TV broadcast signal, a playback signal such as a DVD or VTR by the playback device, and a signal from the TV broadcast receiving device. A signal input to the input terminal 501 is sent to the reception processing unit 502.

  The reception processing unit 502 performs processing for converting a signal input through the input terminal 501 into a format that can be processed by the gradation conversion unit 500 in the subsequent stage. For example, when an analog signal is input, the reception processing unit 502 performs input signal processing such as processing to convert the analog signal into a digital signal and processing to separate the synchronization signal. For example, when receiving MPEG data, the reception processing unit 502 performs processing such as decoding the MPEG data. The reception processing unit 502 outputs the processed image signal to the gradation conversion unit 500.

  The gradation conversion means 500 performs gradation conversion processing similar to any one of the gradation conversion means 1 to 5 in the image processing apparatuses of Embodiments 1 to 5 described above, and outputs an image signal having multiple gradations than the input image signal. Output to the display means 504.

  The display unit 504 receives the image signal Dc whose number of bits is expanded from the gradation conversion unit 500, and displays an image with a displayable number of bits based on the received image signal.

  As described above, according to the image display device 6 of the sixth embodiment, the image processing device or the image processing method of any of the first to fifth embodiments is applied to the gradation conversion means 500. When the number of bits of the input image signal is smaller than the number of bits that can be displayed by the display unit 504, or when the display device (for example, a liquid crystal device, a plasma display, etc.) provided in the display unit 504 is a high-luminance device. In addition, even when the gradation value changes very slowly by real-time processing without greatly increasing the circuit scale, the change in gradation value is detected more appropriately without erroneous detection due to noise components. Thus, the gradation of the image signal can be expanded to a sufficient number of gradations. For this reason, according to the image display device 6 of the sixth embodiment, even when an image is displayed on a display device with high brightness, the quantization of the image is not lost and the sharpness is not lost. It is possible to display an image signal in which noise and pseudo contour are reduced, the gradation value changes smoothly, and the number of gradations is extended with high quality.

<< 7 >> Embodiment 7
The image display device 6 according to the sixth embodiment includes one gradation conversion unit 500, but the image display device 7 according to the seventh embodiment includes a plurality of gradation conversion units 500 and 510. Yes. FIG. 40 is a block diagram schematically showing an example of the configuration of the image display device 7 according to the seventh embodiment of the present invention. 40, the same reference numerals are given to the same or corresponding components as those in FIG. In FIG. 40, the image display device 7 of the seventh embodiment is further provided with a tone correction unit 503 and a tone conversion unit 510 as signal processing units, and the image display device 6 of the sixth embodiment. Is different. The gradation correction unit 503 is a unit that performs gradation correction such as expansion of the dynamic range of the image signal and improvement of contrast. The gradation conversion unit 510 performs gradation conversion processing similar to the gradation conversion units 1 to 5 in the first to fifth embodiments, and outputs an image signal having a higher gradation than the input image signal to the display unit 504. To do. The gradation conversion unit 510 performs quantization correction such as expansion of the dynamic range of the image signal and improvement of contrast by the gradation correction unit 503, and the quantization noise of the image signal increased with the gradation correction. In order to reduce (gradation jump), gradation conversion processing is performed on the image signal after gradation correction.

  An image signal input from the input terminal 501 is received by the reception processing unit 502, and an image signal in a predetermined format is output to the gradation conversion unit 500. The gradation conversion unit 500 performs gradation conversion processing for expanding the number of bits on the input image signal to reduce quantization noise. The configuration and operation of the seventh embodiment so far are the same as those in the sixth embodiment.

  The gradation correction unit 503 receives an image signal from the gradation conversion unit 500 that is an image signal having a higher gradation than the input image signal. The gradation correction unit 503 performs gradation correction such as expansion of the dynamic range of the input image signal and improvement of contrast. In the gradation correction, a process for converting the gradation of a certain section of the input image signal at a predetermined magnification is performed. As a result, the dynamic range of the displayed image is expanded, and an image having a clear contrast and high contrast is displayed. However, the gradation correction performed by the gradation correcting unit 503 makes it easy for a gradation jump to occur. , Resulting in increased quantization noise. That is, for example, when correction is made such that the gradation value is quadrupled, the gradation of the output image has only values every four gradations, and even in an area where the image gradually becomes brighter, The gradation value changes every four gradations, and this gradation jump becomes conspicuous as quantization noise.

  The image signal output from the gradation correction unit 503 is input to the gradation conversion unit 510. The gradation conversion unit 510 performs gradation conversion processing for expanding the number of bits on the input image signal by the same processing as the gradation conversion unit 500, and performs processing so that the change in gradation value in the image signal becomes smooth. Then, the obtained image signal is output to the display means 504.

  Here, in the gradation correction unit 503, for example, when processing that causes a gradation jump of 4 gradations is performed, the gradation conversion unit 510 is set as described in the first to fifth embodiments. What is necessary is just to set so that the difference for four gradations, that is, the change of the gradation value in the range from +4 to -4 can be detected. In this way, the change value of the gradation value detected by the gradation converting means 510 may be determined according to the characteristics of the gradation correcting means 503. In this case, the configuration and operation of the gradation conversion unit 510 are the same as those of any of the gradation conversion units 1 to 5 in the first to fifth embodiments.

  The display unit 504 receives an image signal in which the number of bits has been expanded by processing the gradation value change from the gradation conversion unit 510 to be smooth, and the number of bits that can be displayed based on the received image signal To display the image.

  As described above, according to the image display device 7 of the seventh embodiment, after the quantization noise is reduced by the gradation conversion unit 500, the dynamic range expansion and the contrast correction processing are performed by the gradation correction unit 503. Even in this case, the gradation jump is generated by setting the change value (maximum value, minimum value) of the gradation value detected by the gradation conversion means 510 according to the characteristics of the gradation correction means 503. It is possible to perform gradation conversion processing for reducing gradation jumps without searching for gradation values. Therefore, according to the image display device 7 of the seventh embodiment, the circuit scale is greatly increased even when the display device (for example, a liquid crystal device, a plasma display, etc.) provided in the display unit 504 is a high-luminance device. Therefore, even in a section where the gradation value changes very slowly by real-time processing, the gradation change of the image signal is detected more appropriately without erroneous detection due to noise components. Even when the image is displayed on a display device with high brightness that can be expanded to a sufficient number of gradations, the quantization noise and pseudo contours are maintained without losing image details or loss of sharpness. Can be displayed, and the change of the gradation value is smooth, and an image signal with an extended number of gradations can be displayed.

  FIG. 40 shows a case where the gradation correction unit 503 for performing gradation correction is arranged at the subsequent stage of the gradation conversion unit 500. However, instead of the gradation correction unit 503, signal processing other than gradation correction is performed. Even when a means (not shown) for increasing quantization noise due to gradation jump is provided, the gradation conversion means 510 can reduce the quantization noise. In this case, the change value (maximum value, minimum value) of the gradation value detected by the gradation conversion means 510 is increased by image processing if it is determined according to the quantization noise characteristics of the signal processing means. It can be configured to reduce quantization noise.

  1, 2, 3, 4, 5 gradation conversion means, 6, 7 image display device, 10 gradation change detection means, 11 gradation change calculation means, 12 gradation change restriction means, 20, 20b, 70 , 70b pixel area determination means, 21, 21b pixel change amount extraction means, 22 absolute value level comparison means, 23, 24, 27, 28 adjacent position polarity determination means, 25, 29 synthesis means, 26 area signal generation means, 30 flat Noise area width determination means, 31 flat / noise area width count means, 32 cyclic filter processing means, 33 coefficient 1 multiplication means, 34 calculation means, 35 coefficient 2 multiplication means, 36 switching means, 37 filter output delay means, 40 , 41 Inclination detecting means, 40a Inclination calculating means, 40b Switching means, 40c Inclination delay means, 42 End point change amount detection Means, 43 inclination calculating means, 50, 81, 81b correction bit generating means, 51 calculating means, 52 subtracting means, 53 switching means, 54, 84 correction value delay means, 55, 85 bit string converting means, 60 edge change detecting means, 61 absolute value calculation means, 62 differential low frequency component extraction means, 63 absolute value calculation means, 64 magnification conversion means, 65 maximum value selection means, 66 coefficient conversion means, 67 region signal delay compensation means, 72, 76 smoothing means, 73 polarity determination means, 74 comparison means, 75 area signal generation means, 80 flatness / noise area width determination means, 82 addition means, 83 switching means, 86 count means, 87 multiplication means, 91, 93 delay compensation means, 92, 94 Pixel value calculation means, 110 horizontal gradation processing means, 111 vertical direction Direction gradation processing means, 500, 510 gradation conversion means, 501 input terminal, 502 reception processing means, 503 gradation correction means, 504 display means, Ar, Ar1, Ar2, Ar3, Ar4 area signal, aCd correction value, Cd , Ce correction bit data, Cnt flat / noise area width count value, Da input image signal (input pixel data), Db output image signal (output pixel data), Dc output image signal (output pixel data), df gradation change amount , De flat / noise area width data measured value, dst flat / noise area width data predicted value, E additional bit data, Ksl, Ks slope data value, La (x) tone value of input image signal, Lb (x) output Image signal tone value, rc correction limit coefficient, x pixel position, Xlm amplitude limit tone change amount.

Claims (21)

An input image signal composed of discrete data in the time direction and the spatial direction is inputted, and composed of post-conversion data generated by a conversion process that expands the number of gradations of the input image signal and corrects the gradation value. An image processing apparatus having a function of outputting an output image signal,
A gradation change amount detecting means for obtaining a gradation change amount which is a difference between gradation values of pixels arranged adjacent to each other from the input image signal;
Based on the gradation change amount, the value of the gradation change amount is determined by the pixels belonging to the gradation flat region where the gradation change amount value is 0 and the noise component included in the input image signal. A region signal indicating a result of the determination by classifying the pixel into a pixel belonging to a non-zero noise region and a pixel in a non-noise gradation change position that is a region other than the gradation flat region and the noise region. A pixel region determination means for generating
The width of the flat / noise area consisting of the gradation flat area and the noise area between the reference pixel which is the pixel at the non-noise gradation change position and the next pixel at the next non-noise gradation change position is received. A means for generating a flat / noise region width data prediction value indicating a predicted value of a certain flat / noise region width, wherein the flat / noise region width immediately before the reference pixel and the immediately preceding reference detected immediately before the reference pixel A flat / noise area for generating the flat / noise area width data predicted value obtained by predicting the width of the flat / noise area immediately after the reference pixel based on the predicted flat / noise area width data immediately before generated in the pixel. Width determination means;
Inclination data value used to set the gradient of the gradation value of the converted data in the flat / noise region immediately after the reference pixel based on the gradation change amount and the predicted flat / noise region width data value Inclination detecting means for obtaining
Using the gradation change amount and the inclination data value, a correction value of the gradation value of the input image signal corresponding to the determination result indicated by the region signal is obtained, and correction bit data corresponding to the correction value is obtained. Correction bit generation means for generating,
A pixel value calculation means for subtracting or adding a value corresponding to the correction bit data from the gradation value to generate the output image signal composed of the converted data. Processing equipment. The flat / noise area width determining means is:
Flat / noise area width count means for receiving the area signal and outputting a count value of the number of pixels of the flat / noise area width as a flat / noise area width count value indicating the flat / noise area width;
For each of the reference pixels, the flat / noise area width count value is obtained up to the pixel immediately before the reference pixel by applying a cyclic filter process to the filter processing output up to the pixel immediately before the reference pixel. 2. A cyclic filter processing means for obtaining a low frequency component of a flat / noise region width value and outputting the obtained value as the flat / noise region width data prediction value. The image processing apparatus described. The inclination detecting means is
When setting the gradient of the gradation value of the converted data based on the amplitude-limited gradation variation amount in which the gradation variation value is limited to a predetermined range and the flat / noise area width data prediction value An inclination calculating means for obtaining an inclination used for
For each reference pixel, the gradient of the gradation value obtained by the gradient calculation means is selected as the gradient data value, and the gradient value of the previous pixel is used as the gradient data value for pixels other than the reference pixel. Switching means to select,
3. The image according to claim 1, wherein an output from the switching unit is output as the inclination data value indicating an inclination of a gradation value of the image signal in a section corresponding to the flat / noise area. Processing equipment. Edge change detection means for setting a correction limiting coefficient for limiting the gradient of the gradation value of the output image signal based on the gradation change amount;
The correction bit generation means uses the gradation change amount, the inclination data value, and the correction restriction coefficient to calculate a correction value of the gradation value of the input image signal according to the determination result indicated by the region signal. The image processing apparatus according to claim 1, wherein the image processing apparatus obtains and generates correction bit data corresponding to the correction value. The edge change detecting means is
If the absolute value of the gradation change amount is larger than a predetermined threshold value, it is determined that the gradation change amount value is due to an edge of the image, the correction limit coefficient value is set to 0,
When the absolute value of the gradation change amount is equal to or less than the threshold value, the value of the correction limiting coefficient is set to a value within 1 and less than 0 according to the absolute value of the gradation change amount. The image processing apparatus according to claim 4, wherein: The correction bit generation means includes
An arithmetic means for obtaining a correction value for the gradation value of the input image signal from the amplitude-limited gradation change amount and the correction restriction coefficient in which the value of the gradation change amount is limited to a predetermined range;
Subtraction means for obtaining the inclination data value as a value of change for each pixel of the correction value, subtracting the inclination data value from the correction value in the immediately preceding pixel, and outputting the subtraction result;
Based on the determination result indicated by the area signal, the correction value obtained by the calculation means is selected for a reference pixel that is a pixel at a non-noise gradation change position, and the gradation flat area and noise immediately after the reference pixel are selected. In the pixel to be an area, the subtraction result by the subtraction means is selected, and the switching means for outputting the selected result as a correction value for the gradation value of the input image signal;
Correction bit data having a quantization value corresponding to a fractional part with respect to the gradation value in the input image signal, obtained from the correction value at the reference pixel, by sequentially obtaining the inclination data value for each pixel. The image processing apparatus according to claim 4, further comprising: a bit string conversion unit that generates An input image signal composed of discrete data in the time direction and the spatial direction is inputted, and composed of post-conversion data generated by a conversion process that expands the number of gradations of the input image signal and corrects the gradation value. An image processing apparatus having a function of outputting an output image signal,
A gradation change amount detecting means for obtaining a gradation change amount which is a difference between gradation values of pixels arranged adjacent to each other from the input image signal;
Based on the gradation change amount, the value of the gradation change amount is determined by the pixels belonging to the gradation flat region where the gradation change amount value is 0 and the noise component included in the input image signal. A region signal indicating a result of the determination by classifying the pixel into a pixel belonging to a non-zero noise region and a pixel in a non-noise gradation change position that is a region other than the gradation flat region and the noise region. A pixel region determination means for generating
The width of the flat / noise area consisting of the gradation flat area and the noise area between the reference pixel which is the pixel at the non-noise gradation change position and the next pixel at the next non-noise gradation change position is received. A flat / noise area width determining means for measuring a flat / noise area width indicating a certain flat / noise area width and outputting a flat / noise area width measurement value;
Based on the gradation change amount and the flat / noise area width measurement value, the inclination data value used when setting the inclination of the gradation value of the converted data in the flat / noise area immediately after the reference pixel is A desired tilt detection means;
Using the gradation change amount and the inclination data value, a correction value of the gradation value of the input image signal corresponding to the determination result indicated by the region signal is obtained, and correction bit data corresponding to the correction value is obtained. Correction bit generation means for generating,
Image value processing means comprising: pixel value calculation means for subtracting or adding a value corresponding to the correction bit data from the gradation value to generate the output image signal composed of the converted data. apparatus. The correction bit generation means includes
The inclination data value from the inclination detecting means is set as a value of change for each pixel in the correction value,
According to the determination result indicated by the region signal, the reference value that is the pixel at the non-noise gradation change position is set to a fixed value,
At the pixel position that becomes the gradation flat region and the noise region immediately after the reference pixel, a value obtained by sequentially adding a value corresponding to the inclination data value in the pixel order to the correction value at the reference pixel is used. The image processing apparatus according to claim 7, wherein correction bit data generated by adding as a value corresponding to a decimal part of a gradation value in an input image signal is generated.   The pixel area determination means includes a value of the gradation change amount at the target pixel position, and a polarity of the gradation change amount of each pixel within a predetermined pixel range including the target pixel and pixels positioned before and after the target pixel. The image processing apparatus according to claim 1, wherein the noise area is determined from the image processing apparatus. The pixel area determination means includes
Absolute value level comparing means for comparing the value of the gradation change amount at the target pixel position with a predetermined value and outputting a comparison result;
The gradation change amount in the predetermined pixel range including the target pixel and the pixels positioned before and after the target pixel is obtained, and the polarity of the gradation change amount at the target pixel position and the gradation in the pixel in the predetermined pixel range are obtained. Polarity determining means for determining whether or not the polarity of each of the tone change amounts matches or does not match;
In the determination result in the polarity determination means, a combination means for obtaining a case where the polarity is opposite in any of the pixels within the predetermined pixel range and outputting the polarity determination result;
An area signal generating means for determining a change in gradation to be the noise area based on a comparison result by the absolute value level comparing means and a polarity determination result from the synthesizing means and generating an area signal as a determination result The image processing apparatus according to claim 1, further comprising: The pixel area determination means includes
(Decision A1) When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0, the target pixel is determined to be a pixel belonging to the flat gradation region. ,
(Decision A2) The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd (Nd is a predetermined integer), and the gradation change amount df ( the polarity of x) is opposite to the polarity of the gradation change amount df (x-1), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( If the polarity is opposite to x + 1),
The target pixel is determined to be a pixel belonging to a noise region (determination A3). In the determination A1, it is determined that the target pixel is not a pixel belonging to a flat gradation region, and in the determination A2, the target pixel The image according to any one of claims 1 to 10, wherein the pixel of interest determined that the pixel is not a pixel belonging to a noise region is determined to be a pixel at a non-noise gradation change position. Processing equipment. The pixel area determination means includes
(Determination B1) When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0, it is determined that the target pixel is a pixel belonging to the flat gradation region. ,
(Decision B2) The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd (Nd is a predetermined integer), and the gradation change amount df ( the polarity of x) is opposite to the polarity of the gradation change amount df (x−1), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( x + 1) is opposite to the polarity, or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( when the polarity is opposite to the polarity of x-2), or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the polarity of the gradation change amount df (x) is the gradation change amount df ( When the polarity is opposite to the polarity of x + 2),
The target pixel is determined to be a pixel belonging to the noise region,
(Decision B3) When it is determined in the determination B1 that the pixel of interest is not a pixel belonging to the flat gradation region, and it is determined in the determination B2 that the pixel of interest is not a pixel belonging to the noise region The image processing apparatus according to claim 1, wherein the target pixel is determined to be a pixel at a non-noise gradation change position.   The pixel area determination unit smoothes the gradation change amount in a predetermined pixel range including the value of the gradation change amount at the target pixel position and the target pixel and the pixels positioned before and after the target pixel. The image processing apparatus according to claim 1, wherein a change in gradation that becomes the noise region is determined. The pixel area determination means includes
Absolute value level comparing means for comparing the value of the gradation change amount at the target pixel position with a predetermined value and outputting a comparison result;
Obtain the above-mentioned gradation change amount in a predetermined pixel range including the pixel of interest and the pixels located before and after it, smooth the gradation change amount in the pixels within the predetermined pixel range, and output the smoothed result Smoothing means,
Polarity determination means for determining whether the gradation change amount at the target pixel position matches the polarity of the smoothing result;
A comparison means for comparing the smoothing result with a predetermined value and outputting the comparison result;
Based on the comparison result by the absolute value level comparison unit, the polarity determination result from the polarity determination unit, and the comparison result by the comparison unit, a change in gradation that becomes the noise region is determined, and the region signal is The image processing apparatus according to claim 1, further comprising: a region signal generation unit that generates the image signal. The region signal generating means includes
(Decision C1) When the gradation change amount df (x) in the target pixel at the pixel position x is df (x) = 0, it is determined that the target pixel is a pixel belonging to the gradation flat region,
(Determination C2) The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd (Nd is a predetermined integer), and the polarity of the smoothing result And the polarity of the gradation change amount df (x) are opposite to each other, or
The gradation change amount df (x) in the target pixel at the pixel position x is 0 <| df (x) | ≦ Nd, and the absolute value | ct0 | of the smoothing result is | ct0 | ≦ cth0 ( cth0 is a predetermined threshold)
The target pixel is determined to be a pixel belonging to the noise region,
(Determination C3) When it is determined in the determination C1 that the pixel of interest is not a pixel belonging to the flat gradation region, and it is determined in the determination C2 that the pixel of interest is not a pixel that belongs to the noise region The image processing apparatus according to claim 14, wherein the target pixel is determined to be a pixel at a non-noise gradation change position. The gradation change amount detecting means is
A gradation change amount calculating means for calculating a difference in gradation value between adjacent pixels in a fixed direction in either the horizontal direction or the vertical direction of the image signal, and outputting the difference as the gradation change amount;
16. A gradation change amount limiting means for limiting a range in which the value of the gradation change amount can be taken and outputting an amplitude-limited gradation change amount with a limited value. The image processing apparatus according to claim 1.   The gradation change amount limiting means limits the value of the gradation change amount within a range from a maximum value that is a predetermined positive value to a minimum value that is a predetermined negative value, The image processing apparatus according to claim 16, wherein the image processing apparatus outputs the amplitude-limited gradation change amount. An input image signal composed of discrete data in the time direction and the spatial direction is inputted, and composed of post-conversion data generated by a conversion process that expands the number of gradations of the input image signal and corrects the gradation value. An image processing method for outputting an output image signal,
A gradation change amount detection step for obtaining a difference in gradation value between pixels in the spatial direction of the image signal as a gradation change amount when the gradation changes;
Based on the gradation change amount, the value of the gradation change amount is determined by the pixels belonging to the gradation flat region where the gradation change amount value is 0 and the noise component included in the input image signal. A determination is made to classify the pixel into a pixel belonging to a non-zero noise region, and a pixel at a non-noise gradation change position other than the gradation flat region and the noise region, and an area signal indicating the result of the determination is generated A pixel region determination step;
Based on the above area signal, the width of the flat / noise area consisting of the gradation flat area and the noise area between the reference pixel that is the pixel at the non-noise gradation change position and the pixel at the next non-noise gradation change position A means for generating a flat / noise area width data prediction value indicating a predicted value of the flat / noise area width, and immediately before the flat / noise area width immediately before the reference pixel and detected immediately before the reference pixel. The flat noise generating the flat / noise area width data predicted value obtained by predicting the width of the flat / noise area immediately after the reference pixel based on the predicted flat / noise area width data immediately before the reference pixel. An area width determination step;
Inclination data value used to set the gradient of the gradation value of the converted data in the flat / noise region immediately after the reference pixel based on the gradation change amount and the predicted flat / noise region width data value An inclination detection step for obtaining
Using the gradation change amount and the inclination data value, a correction value of the gradation value of the input image signal corresponding to the determination result indicated by the region signal is obtained, and correction bit data corresponding to the correction value is obtained. A correction bit generation step to be generated;
A pixel value calculating step of subtracting or adding a value corresponding to the correction bit data from the gradation value to generate the output image signal composed of the converted data. Method. An input image signal composed of discrete data in the time direction and the spatial direction is inputted, and composed of post-conversion data generated by a conversion process that expands the number of gradations of the input image signal and corrects the gradation value. An image processing method for outputting an output image signal,
A gradation change amount detection step for obtaining a difference in gradation value between pixels in the spatial direction of the image signal as a gradation change amount when the gradation changes;
Based on the gradation change amount, the value of the gradation change amount is determined by the pixels belonging to the gradation flat region where the gradation change amount value is 0 and the noise component included in the input image signal. A determination is made to classify the pixel into a pixel belonging to a non-zero noise region, and a pixel at a non-noise gradation change position other than the gradation flat region and the noise region, and an area signal indicating the result of the determination is generated A pixel region determination step;
Based on the above area signal, the width of the flat / noise area composed of the gradation flat area and the noise area between the reference pixel that is the pixel at the non-noise gradation change position and the pixel at the next non-noise gradation change position A flat / noise area width determination step of measuring a flat / noise area width indicating a flat / noise area width and outputting a flat / noise area width measurement value;
Based on the gradation change amount and the flat / noise area width measurement value, the inclination data value used when setting the inclination of the gradation value of the converted data in the flat / noise area immediately after the reference pixel is A desired tilt detection step;
A correction bit generation step of obtaining a correction value of the gradation value of the input image signal using the gradation change amount and the inclination data value, and generating correction bit data corresponding to the correction value;
A pixel value calculating step of subtracting or adding a value corresponding to the correction bit data from the gradation value to generate the output image signal composed of the converted data. Method. An image processing apparatus according to any one of claims 1 to 17,
An image display device comprising: display means for displaying an image based on the output image signal in which the number of gradations output from the image processing device is expanded. A first image processing unit having the same configuration as that of the image processing apparatus according to any one of claims 1 to 17,
Signal processing means for processing a signal output from the first image processing means;
A second image processing unit having the same configuration as that of the image processing apparatus according to any one of claims 1 to 17, which performs gradation conversion processing on a signal output from the signal processing unit;
An image display apparatus comprising: display means for displaying an image based on an image signal with an expanded number of gradations output from the second image processing means.

Images