JP5312171B2 - Image processing apparatus and method - Google Patents

Description

  The present invention relates to an image processing apparatus and method.

  Conventionally, there is an image processing method for executing gradation conversion processing (for example, γ conversion processing) in accordance with the characteristics of an input image signal (hereinafter, such γ conversion processing is referred to as dynamic γ processing). Specifically, a luminance histogram indicating the characteristics of the input image (histogram representing the distribution of the number of pixels (frequency: frequency) present in the image for each gradation) is created, and dynamic γ processing is performed based on the luminance histogram. Done. The created luminance histogram greatly affects image quality correction (gradation conversion processing) such as dynamic γ processing.

  The input video signal may include pixels that are locally darker than the surroundings (local dark portion pixels). For example, many local dark area pixels exist in an edge portion such as a character area or an OSD (On Screen Display). If a luminance histogram is created including a local dark area pixel when generating a gamma curve in a dark area, the created luminance histogram becomes a histogram in which a large number of frequencies are distributed in dark gradations. In such a gradation conversion process according to the luminance histogram, a gradation correction parameter (a γ curve in the γ conversion process) is generated that floats the gradation of the character area and the edge of the OSD.

  Conventional techniques in view of such problems are disclosed in, for example, Patent Documents 1 and 2. Specifically, in the technique disclosed in Patent Document 1, the luminance value and the color difference value are given a certain range, and the pixels within the ranges are used as the pixels constituting the character area. Then, a luminance histogram is generated from pixels that are not the character area, and the pixels constituting the character area are excluded. In the technique disclosed in Patent Document 2, the input luminance signal is divided into a high frequency component and a low frequency component, and an edge histogram is created for the high frequency component. Then, based on the edge histogram and the luminance histogram, it is determined in which gradation range the gradation correction characteristic is sharpened.

JP 2005-108208 A JP 2006-33469 A

  However, since the technique disclosed in Patent Document 1 uses only pixels that are determined not to be character regions, a brightness histogram different from the actual brightness histogram is created (the brightness histogram is processed). For this reason, if such a luminance histogram is used in image processing other than dynamic γ processing, an influence is caused. That is, the luminance histogram is often used for image processing other than gradation correction, but the luminance histogram created by the technique disclosed in Patent Document 1 cannot be used for other image processing. As a result, a luminance histogram that can be used in other image processing is separately generated, and it is considered that the image processing apparatus is wasted. Furthermore, the technique disclosed in Patent Document 1 cannot detect a local dark pixel. Further, in the technique disclosed in Patent Document 2, the edge and luminance distribution of the entire screen is detected, but whether or not the pixel in the edge portion is locally darker than the surrounding pixels, that is, the pixel is locally It cannot be determined whether the pixel is a dark pixel. Therefore, it is not possible to generate a tone correction parameter that takes into account local dark pixels.

  SUMMARY OF THE INVENTION An object of the present invention is to provide an image processing apparatus and method that can generate tone correction parameters that are not affected by local dark pixels, and that can improve tone reproducibility.

  In order to solve the above-described problem, an image processing apparatus according to the present invention includes a creation unit that creates a brightness histogram from brightness values for each pixel of an input image, and a brightness value for each pixel of the image. Determination means for determining whether or not the pixel is a locally dark local pixel from the luminance values of the surrounding pixels, and a gradation correction parameter for correcting the gradation of the image, the luminance histogram and the Generating means using the total number of local dark area pixels, and the generating means, when the total number of local dark area pixels is equal to or greater than a predetermined threshold, the gradation of the local dark area pixels from the luminance histogram Detecting a tone of interest having a maximum frequency from a range excluding a luminance region including the tone region, and generating a tone correction parameter for increasing a tone property in the vicinity of the tone of interest including the tone of interest. To do.

The image processing method of the present invention includes a creation step of creating a luminance histogram from the luminance value for each pixel of the input image, and the luminance value of the pixel and the luminance value of the surrounding pixels for each pixel of the image. A determination step for determining whether the pixel is a locally dark local dark pixel, a gradation correction parameter for correcting the gradation of the image, the luminance histogram and the total number of the local dark pixel. a generation step of generating with, the said at generating step, when the total number of the local dark portion pixel is not smaller than a predetermined threshold value, except for the luminance region including the gradation of the local dark portion pixel from the luminance histogram It is characterized in that a target gradation having a maximum frequency is detected from the range, and a gradation correction parameter for increasing the gradation in the vicinity of the target gradation including the target gradation is generated.

  According to the present invention, it is possible to provide an image processing apparatus and method that can generate gradation correction parameters that are not affected by local dark pixels, and that can improve gradation reproducibility.

It is a functional block diagram of the image processing apparatus which concerns on this embodiment. It is a figure which shows an example of a division area. It is a figure which shows an example of the filter used for determination of a local dark part pixel. It is a figure which shows an example of the flow of a judgment process of a local dark part pixel. It is a figure which shows an example of an input image. It is a figure which shows an example of the brightness | luminance histogram of a division area. It is a figure which shows an example of the flow of an attention gradation determination process. It is a figure which shows an example of the gradation correction parameter produced | generated.

  Hereinafter, an image processing apparatus and method according to embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a functional block diagram of an image processing apparatus 100 according to the present embodiment. The image processing apparatus 100 includes a luminance calculation unit 101, a luminance histogram creation unit 102, a divided region information generation unit 103, a local dark part pixel determination unit 104, a local dark part pixel count unit 105, an attention gradation detection unit 106, and an image quality correction unit 107. Have. In the present embodiment, a configuration for correcting the gradation of a dark part (for example, half or less of the maximum gradation) will be described.

The luminance calculation unit 101 obtains a luminance value (luminance data) for each pixel of the input image (input image). When the input image data (input image signal) is RGB image data, color matrix calculation is performed to obtain luminance data.

  The luminance histogram creation unit 102 creates a luminance histogram from the luminance value for each pixel of the input image. In the present embodiment, the luminance histogram creation unit 102 divides the input image into a plurality of regions (divided regions), and creates a luminance histogram for each divided region from the luminance data obtained by the luminance calculation unit 101. For example, the input image is a 1920 × 1080 image. Then, as shown in FIG. 2, the input image is divided into six (3) × 2 (row) divided regions, and a luminance histogram for each divided region is created. In the present embodiment, in the luminance histogram, it is assumed that the gradation is divided into 128 categories from 0 gradation representing black to 127 gradation representing white. Note that the number of divided areas is not limited to 3 × 2. For example, it may be 10 × 5 or 5 × 3, or may not be divided. Further, the number of brightness histogram categories is not limited to 128. For example, 64 or 256 may be used.

  Note that the divided region information representing each divided region is generated by the divided region information generation unit 103 and sent to the luminance histogram generation unit 102 and a local dark portion pixel count unit 105 described later. Further, as shown in FIG. 2, it is assumed that numbers blk00, blk10, blk20, blk01, blk11, blk21 (identifier) are associated with each divided region.

  For each pixel of the input image, the local dark part pixel determination unit 104 determines whether the target pixel is a local dark part pixel that is locally dark from the brightness value of the pixel (target pixel) and the brightness value of the surrounding pixels (peripheral pixels). Judge whether there is. The local dark part pixel determination unit 104 will be described in more detail with reference to FIGS. FIG. 3 is a diagram illustrating the shape of a filter for determining whether or not a target pixel is a local dark pixel. In FIG. 3, a circle corresponds to luminance data for one pixel. FIG. 4 is a flowchart illustrating an example of a processing flow of the image processing apparatus when it is determined whether the target pixel is a local dark portion pixel.

  In step S401, the local dark part pixel determination unit 104 initializes the local dark part pixel determination flag usth_f. In the present embodiment, the initial value of the local dark area pixel determination flag usth_f is set to 0.

  In step S402, the local dark pixel determination unit 104 compares the luminance data b (x, y) of the target pixel (the pixel (x, y) illustrated in FIG. 3) with a predetermined luminance threshold usth. If the luminance data of the target pixel is smaller than the luminance threshold usth (step S402: YES), it is determined that the target pixel is a dark pixel, and the process proceeds to step S403. If the luminance data of the pixel of interest is equal to or greater than the luminance threshold usth (step S402: NO), it is determined that the pixel of interest is not a dark pixel, and the process ends.

  In step S403, the local dark area pixel determination unit 104 determines from the luminance data of the peripheral pixels of the target pixel (x, y) (for example, the luminance data b (x-2, y) of the pixel (x-2, y)). The luminance data b (x, y) of the target pixel (x, y) is subtracted. The peripheral pixels are, for example, the pixels (x, y-2), (x, y-1), (x, y + 1), (x, y + 2), (x-2, y), (x-) shown in FIG. 1, y), (x + 1, y), (x + 2, y). Then, it is obtained by subtracting the luminance data of the pixel of interest (x, y) from the luminance data of the pixels (x, y-2), (x, y-1), (x, y + 1), (x, y + 2). The difference values to be obtained are referred to as difference values diff0 to diff3, respectively. The difference obtained by subtracting the luminance data of the pixel of interest (x, y) from the luminance data of the pixels (x-2, y), (x-1, y), (x + 1, y), (x + 2, y). The values are set as difference values diff4 to diff7, respectively.

In step S404, the local dark pixel determination unit 104 compares the difference values diff0 to diff7 obtained in step S403 with a predetermined difference threshold value. Difference value diff0
When at least one of ~ diff7 is larger than the difference threshold (step S404: YES), it is determined that the target pixel is a local dark pixel, and the process proceeds to step S405. If all of the difference values diff0 to diff7 are equal to or less than the difference threshold value (step S404: NO), it is determined that the pixel of interest is not a local dark pixel, and the process ends. That is, in this embodiment, the luminance value of the target pixel is smaller than a predetermined luminance threshold value, and the difference between the luminance value of the target pixel and the luminance value of at least one peripheral pixel is larger than the predetermined difference threshold value. In addition, the target pixel is a local dark pixel.

  In step S405, the local dark part pixel determination unit 104 sets the local dark part pixel determination flag usth_f to 1 and ends. That is, the local dark area pixel determination flag usth_f is set to 1 for the local dark area pixels, and the local dark area pixel determination flag usth_f is set to 0 for the other pixels. The processing described above is performed on each pixel for each divided region.

  The local dark part pixel count unit 105 counts the number of local dark part pixel determination flags usth_f set to 1 in the local dark part pixel determination unit 104, that is, the number of local dark part pixels for each divided region. Then, the count result (total number of local dark area pixels) is generated for each divided region as the local dark area pixel number yheu.

  The target gradation detection unit 106 and the image quality correction unit 107 generate gradation correction parameters for correcting the gradation of the input image using the luminance histogram and the number of local dark part pixels. The gradation correction parameter is a curve (gradation conversion curve), a table, or a function in which an input gradation and an output gradation are associated with each other. In this embodiment, a gradation conversion curve (γ Curve). Hereinafter, functions of the target gradation detection unit 106 and the image quality correction unit 107 will be described in detail.

  The attention gradation detection unit 106 detects the attention gradation for each divided region based on the luminance histogram created by the luminance histogram creation unit 102 and the local dark part pixel count yheu obtained by the local dark part pixel count part 105. Note that the target gradation is a main gradation value (or gradation range) of a subject to be noted (target subject) in the image. The attention gradation detecting unit 106 will be described in more detail with reference to FIGS.

  In the present embodiment, description will be made assuming that an input image having a size of 1920 × 1080 as shown in FIG. 5 is input. In the divided region blk01, as shown in FIG. 5, it is assumed that a star-shaped line 501 corresponding to a local dark portion pixel and a subject of interest 502 darker than the intermediate color exist in the intermediate color background. Further, the local dark portion pixel number yheu in the region blk01 is 45.

  FIG. 6 is a luminance histogram of the divided region blk01 shown in FIG. An example of the processing flow of the image processing apparatus when determining the target gradation will be described with reference to the flowchart of FIG.

  First, in step S701, the target gradation detection unit 106 initializes a gradation number i. In the present embodiment, the initial value of the number i is 0, that is, the darkest 0 gradation is the initial gradation. As a result, the target gradation is detected from the darkest part.

In step S702, the target gradation detection unit 106 compares the luminance data d (i) of i gradation and the luminance threshold usth. If the i gradation luminance data is smaller than the luminance threshold usth (step S702: YES), the process proceeds to step S703. If the i gradation luminance data is equal to or greater than the luminance threshold usth (step S702: NO), the process proceeds to step S704. A description will be given by taking the eight gradations in FIG. 6 as an example. Note that the luminance threshold usth corresponds to 10 gradations as shown in FIG. Since 8 gradations correspond to gradations darker than 10 gradations, the process proceeds to step S703.

  In step S <b> 703, the target gradation detection unit 106 compares the local dark part pixel count yheu generated by the local dark part pixel count unit 105 with a predetermined determination threshold eglim. If the local dark pixel number yheu is equal to or greater than the determination threshold eglim (step S703: YES), the process proceeds to step S705. If the local dark pixel number yheu is smaller than the determination threshold eglim (step S703: NO), the process proceeds to step S704.

  In step S704, the target gradation detection unit 106 determines whether or not the i gradation is set as the target gradation (in the dark part) (attention gradation determination process). There are several methods for this determination. In the present embodiment, when the frequency (frequency) of i gradation reaches a maximum, the i gradation is set as the attention gradation. The maximum corresponds to a case where a value larger than the frequency of other peripheral gradations is shown, or a case where the frequency becomes higher than a predetermined threshold. Whether or not the frequency is maximal may be determined, for example, by confirming the frequency of gradation within a predetermined range (for example, i ± 2). Further, in an image including many pixels of the same gradation such as letterbox and black belt, the gradation of the pixel may not be set as the attention gradation. That is, when the total number of local dark area pixels is equal to or greater than the determination threshold, a target gradation having a maximum frequency is detected from a range obtained by excluding the luminance area including the gradation of the local dark area pixels from the luminance histogram.

  In step S705, the target gradation detecting unit 106 searches for the target gradation when the target gradation is determined or until the gradation that is the boundary between the dark part and the bright part (the 63rd gradation in the example of FIG. 6). When the process is performed, the process is terminated. In other cases, the target gradation detecting unit 106 adds 1 to i (step S706), and the process returns to step S702.

  Here, the description will be supplemented by taking the eighth gradation in FIG. 6 as an example. In the present embodiment, the determination threshold eglim is 32. Further, it is assumed that the local dark portion pixel number yheu of blk01 is 45. In that case, in the process in step S703 of FIG. 7, since the local dark portion pixel number yheu is equal to or larger than the determination threshold eglim, step S704 is skipped and the process proceeds to step S705. As a result, the eighth gradation is not the attention gradation (although the gradation has the maximum frequency). If the local dark pixel number yheu is smaller than the determination threshold eglim, the eighth gradation is set as the target gradation. When the local dark portion pixel number yheu is equal to or larger than the determination threshold eglim and a luminance histogram as shown in FIG. 6 is obtained, the 25th gradation is set as the attention gradation instead of the eighth gradation.

  The image quality correction unit 107 generates a gradation correction parameter (γ curve in the present embodiment) based on the dark part attention gradation (dark part attention gradation) determined by the attention gradation detection unit 106. Specifically, a gradation correction parameter that increases the gradation in the vicinity of the target gradation including the target gradation is generated. In the present embodiment, the tone correction parameter is generated for each divided region. Then, tone conversion is performed using the generated (created) γ curve, and γ-converted data (image data) is output. An example of the γ curve created is shown in FIG. In FIG. 8B, a solid line indicates a created γ curve, and a broken line indicates a conversion straight line that directly outputs an input image.

In the region blk01, as shown in FIG. 6, 25 gradations are dark part attention gradations. Therefore, on the low gradation side, as shown in FIG. 8B, a wide range of output gradations (converted gradations) are assigned in the vicinity of 25 gradations (input gradations) detected as dark part attention gradations. A gamma curve is created. Note that the γ curve shown in the figure is an example, and an arbitrary γ curve that increases the gradation in the vicinity of the target gradation may be generated. Increasing the gradation in the vicinity of the target gradation corresponds to increasing the output gradation range in comparison with the input gradation range in a predetermined gradation range including the attention gradation. Therefore, the output gradation is assigned so that the degree of gradation change near the target gradation of the input video is more emphasized. Note that by increasing the gradation of the dark part attention gradation, an image in which the local dark part pixel is relatively emphasized is suppressed, but the level of the input gradation corresponding to the local dark part pixel is more positive. You may make it produce | generate the gamma curve which reduces a tonality. In the γ curve shown in FIG. 8B, in the vicinity of the input gradation value of 8, the output gradation range is smaller than the input gradation range. By doing in this way, it becomes possible to suppress the emphasis of the edge part which corresponds to a local dark part pixel more.

  FIG. 8A shows a gamma curve in a dark part when the present invention is not applied. When the present invention is not applied, a γ curve in which a wide range of output gradations are assigned in the vicinity of 8 gradations is created. As a result, even if the target gradation (25 gradations) exists, an image in which the target gradation is not preferably expressed is output. Specifically, an edge portion corresponding to a local dark portion pixel of a character or OSD is emphasized, and an image with black as a whole is output. However, when the present invention is applied, it is possible to suppress the edge portion corresponding to the local dark pixel from being emphasized and to express the target gradation favorably.

  As described above, according to the image processing apparatus and method according to the present embodiment, the gradation correction parameter is generated using the luminance histogram and the number of local dark area pixels. As a result, tone correction parameters that are not affected by local dark pixels can be generated, and tone reproducibility can be improved. For example, the target gradation can be detected from the range excluding the luminance region including the gradation of the local dark pixel. That is, the gradation that should originally improve the gradation can be set as the attention gradation. Accordingly, it is possible to generate a gradation correction parameter that can improve the gradation of the vicinity of the target gradation including the target gradation without increasing the gradation of the gradation of the edge portion of the character or OSD. Therefore, a scene such as a subtle shadow in a dark part can be reproduced well.

  Further, according to the image processing apparatus according to the present embodiment, a luminance histogram including characters and pixels at the edge portion of the OSD is created. That is, since the luminance histogram is not processed as in the prior art, the luminance histogram does not affect other image processing.

  In the example of FIG. 6, the case where there is one target gradation has been described. However, when there are more gradations having the maximum frequency, these gradations may be used as the target gradation. Then, a gradation correction parameter in which a wide range of output gradations are assigned in the vicinity of all or some of the plurality of attention gradations may be created. When the total number of local dark area pixels is equal to or greater than the determination threshold, it is not necessary to create a gradation correction parameter that increases the gradation of the luminance region including the gradation of the local dark area pixels described above.

  In addition, when a plurality of gradations are set as the attention gradations, the gradation correction that enhances the gradation in the vicinity of the attention gradation including at least the attention gradation on the lowest gradation side among the plurality of attention gradations. It is preferable to generate parameters. If at least such tone correction parameters are generated, tone reproducibility can be improved.

  Note that, since the gradation indicating black differs depending on the type of input image, the luminance threshold usth is preferably different for each type of input image. For example, when the number of gradation histogram categories is 128, black in an image generated by a digital camera or a personal computer corresponds to 0 gradation, and black in an image generated by a television has 8 gradations. Correspond.

In the present embodiment, the tone correction parameter is generated based on the target tone. However, how the tone correction parameter is generated based on the luminance histogram and the total number of local dark pixels. It may be generated. For example, when the number of local dark area pixels is larger than the determination threshold value, a γ curve corresponding to the local dark area pixel number (a γ curve that does not float black) may be generated.

  Note that, in the present embodiment, the target gradation with the highest frequency is detected from the range excluding the luminance region smaller than the luminance threshold, but the range in which the target gradation is detected is not limited to this. The target gradation may be detected from a range obtained by excluding the luminance region including the gradation of the local dark pixel from the luminance histogram. For example, the target gradation may be detected from a range excluding only the gradation of the local dark area pixel or several gradations before and after the gradation of the local dark area pixel (± 3, 5 gradation, etc.).

DESCRIPTION OF SYMBOLS 100 Image processing apparatus 102 Luminance histogram creation part 104 Local dark part pixel judgment part 106 Attention gradation detection part 107 Image quality correction part

Claims (10)

Creating means for creating a luminance histogram from the luminance value of each pixel of the input image;
For each pixel of the image, from the luminance value of the pixel and the luminance value of the surrounding pixels, a determination unit for determining whether the pixel is a locally dark local pixel,
Generating means for generating a gradation correction parameter for correcting the gradation of the image using the luminance histogram and the total number of local dark pixels;
Have
The generating means includes
When the total number of the local dark area pixels is equal to or greater than a predetermined threshold, a target gradation having a maximum frequency is detected from a range excluding the luminance area including the gradation of the local dark area pixels from the luminance histogram,
An image processing apparatus for generating a gradation correction parameter for increasing gradation in the vicinity of an attention gradation including the attention gradation. The generation means generates a gradation correction parameter for increasing the gradation property in the vicinity of the attention gradation including at least the attention gradation on the lowest gradation side when a plurality of the attention gradations are detected. The image processing apparatus according to claim 1, wherein: The determination means, when the luminance value of the pixel is smaller than a predetermined luminance threshold and the difference between the luminance value of the pixel and the luminance value of at least one surrounding pixel is larger than a predetermined difference threshold, The image processing apparatus according to claim 1, wherein the pixel is a local dark pixel. The image processing apparatus according to claim 3, wherein the brightness threshold is different for each type of input image. Further comprising dividing means for dividing the image into a plurality of regions;
The image processing apparatus according to claim 1, wherein the gradation correction parameter is generated for each region. A creation step of creating a brightness histogram from the brightness value of each pixel of the input image;
For each pixel of the image, from the luminance value of the pixel and the luminance value of the surrounding pixels, a determination step of determining whether the pixel is a locally dark local pixel,
The tone correction parameter for correcting the tone of the image, a generation step of generating by using the total number of the luminance histogram and the local dark portion pixel,
Have
In the generating step,
When the total number of the local dark area pixels is equal to or greater than a predetermined threshold, a target gradation having a maximum frequency is detected from a range excluding the luminance area including the gradation of the local dark area pixels from the luminance histogram,
An image processing method for generating a gradation correction parameter for increasing gradation in the vicinity of a target gradation including the target gradation.   In the generation step, when a plurality of the target gradations are detected, a gradation correction parameter that increases the gradation property near the attention gradation including at least the attention gradation on the lowest gradation side is generated.
The image processing method according to claim 6.   In the determination step, when the luminance value of the pixel is smaller than a predetermined luminance threshold and the difference between the luminance value of the pixel and the luminance value of at least one surrounding pixel is larger than a predetermined difference threshold, Make the pixel a local dark pixel
The image processing method according to claim 6 or 7, wherein:   The brightness threshold is different for each type of input image.
The image processing method according to claim 8.   Further comprising a dividing step of dividing the image into a plurality of regions;
  The tone correction parameter is generated for each area.
The image processing method according to any one of claims 6 to 9.

Images