JPH07250326A - Image coding system - Google Patents

Abstract

PURPOSE:To obtain a system converting input image data whose luminance is corrected while taking a characteristic of a display device or the like into account into a code suitable for obtaining a decoded image with less visual deterioration with respect to an original image at a high compression efficiency. CONSTITUTION:The image coder 1 uses a frequency characteristic correction section 11 and an inverse gamma correction section 12 to apply inverse luminance correction to input image data thereby once restoring the data into luminance data before the correction. A control section 13 applies orthogonal transformation or the like to the luminance data and a quantization section 14 applies quantization to a transformation coefficient. Then a variable length coding section 15 and a code generating section 16 are used to code the data and the coded data are sent to a transmission line. An image decoder 2 applies inverse generating processing to the coding to decode the image into an original image and it is outputted on a display device 34.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding method applied to an image communication system and the like, and more particularly to a code capable of reproducing a restored image of image data with respect to original image data with less visual deterioration. Regarding the method of converting to.

[0002]

2. Description of the Related Art Conventionally, of the irreversible encoding techniques for image data, it has been general to employ the configuration shown in FIG. 3 for encoding still images. That is, an image becomes image data by being input from an image input device, for example, a camera 30, but at this time, in order to cancel the non-linear characteristic of the γ characteristic of the display (output device) as shown in FIG. 4A, Normally, the γ correction as shown in FIG. 4B is applied on the camera 30 side. Here, the frequency characteristic may be corrected.

The corrected image data is converted into a predetermined signal in the conversion unit 311 of the image encoding device 31. In the conversion unit 311, in the case of still image coding,
For example, DCT (Discrete Cosine Transf) for each 8 × 8 pixel
orm: Discrete Cosine Transform) is performed, and in the case of moving image coding, regarding frames for which inter-frame prediction is performed,
For example, motion compensation prediction + 8 × 8 DCT is generally performed.

The transform coefficient after the transform is quantized in the quantizer 312. Since quantization is an irreversible process,
Degradation occurs during this process. The quantized data is variable length coded by the variable length coding unit 313. The encoded data obtained in this way is output to the transmission path or the storage device 32 and guided to the image decoding device 33.

On the other hand, in the image decoding device 33, the procedure of the procedure opposite to the above-mentioned encoding is applied to the transmitted encoded data. That is, the variable length coding unit 331 performs variable length decoding, the inverse quantization unit 332 performs inverse quantization, and the inverse conversion unit 333 performs inverse conversion. After that, the output device such as the display 34 applies the γ characteristic and the frequency characteristic as shown in FIG. 4A, and the restored image is displayed on the screen. The output device may be a printer in addition to the display 34.

Among such lossy encoding techniques for image data, the still image encoding system and the moving image encoding system are respectively referred to as "international standard for multimedia encoding".
(Written by Maruzen and Hiroshi Yasuda, 1991)
Techniques of G system and MPEG-VIDEO system are known.

In the encoding / decoding process as described above, the conversion of the conversion unit 311 includes a process of band division of spatial frequency in some sense such as DCT. At that time, since the electric power is concentrated on the low frequency component, and the human visual sensitivity is high in the low frequency range and low in the high frequency range, the quantization unit 312 has a fine low frequency range and a coarse high frequency range. Quantization is performed to reduce unnecessary data. Since the variable-length coding unit 313 performs the variable-length coding after these processes, the quality of the restored image is slightly deteriorated as compared with the case where only the variable-length coding is simply performed, but the code amount is significantly reduced. Is possible.

[0008]

By the way, in the conventional technique, a quantization error is generated in the quantization unit 312 and the dequantization unit 332 during the encoding / decoding. This error is due to the inverse transform unit 333 in the image decoding device 33.
By the processing in (1), noise of the shape of the conversion basis according to the state of the pixel is added to the original image, the noise-containing image data is input to the display 34, and the noise of the output device γ
The characteristics and frequency characteristics are applied and displayed.

Here, paying attention to the generated noise, the noise of the shape of the conversion base in the inverse conversion unit 333 reaches the eyes through the nonlinear system of the γ characteristic and the frequency characteristic of the output device. At this time, the noise caused by the quantization error of the high-pass filter, which should have low visual sensitivity due to the non-linearity of the γ characteristic, contains not only high-frequency components but also DC components and low-frequency components with high visual sensitivity. However, there is a problem that the noise is generated and annoying noise is generated.

The present invention has been made to solve the above problems, and its purpose is to obtain image data with a high compression efficiency and to obtain a restored image with little visual deterioration with respect to the original image. It is to provide a method capable of converting to a different code.

[0011]

The image coding method provided by the present invention is an image coding method for inputting image data for displaying an image by outputting to an output device to generate predetermined coded data. In an image coding method comprising an apparatus and an image decoding apparatus that decodes the coded data into an original image by performing an inverse generation process of the coding, and outputs the decoded data to the output device, The image encoding device performs inverse luminance correction on the input image data, and luminance correction means for generating luminance data simulating the luminance output by the output device, and for the generated luminance data. Transform means for carrying out orthogonal transform or non-orthogonal transform to obtain transform coefficients, quantizing means for quantizing the transform coefficients, and variable length coding for the quantized transform coefficients. Variable-length coding means for generating a variable-length code, code generating means for generating coded data by adding predetermined control information to the generated variable-length code,
It is characterized by including.

In the image coding system having the above-mentioned configuration, the brightness correction means performs, for example, an inverse γ correction unit for performing an inverse γ correction on the input image data, and a brightness correction according to the frequency characteristic of the output device. The quantization means includes at least one of a frequency characteristic correction section, and the quantization means corresponds to a DC component of the conversion coefficient obtained by the conversion means when the luminance correction means includes the inverse γ correction section. It is preferable that the coefficients to be subjected to γ correction be linearly quantized and the other coefficients be linearly quantized as they are.

The present invention also provides an image decoding apparatus for restoring an original image in the image coding system having the above-mentioned configuration. The image decoding device performs a reverse procedure process of the code generating device on the encoded data sent from the image encoding device, and decomposes the variable data into the variable length code and the control information. Variable-length decoding means for performing variable-length code reverse processing of the variable-length coding means to generate variable-length decoded data; and reverse procedure of the quantizing means for the variable-length decoded data. Inverse quantization means for performing processing to generate inverse quantized data, inverse transformation means for subjecting the inverse quantized data to inverse procedure processing of the transformation means, and for data obtained by the inverse transformation means And a reverse brightness correction means for performing reverse procedure processing of the brightness correction means.

In the inverse quantizing means, when the luminance correcting means includes an inverse γ correcting section, the data corresponding to the DC component in the variable length decoded data is subjected to the linear inverse quantization and then the inverse γ is obtained. It is preferable to correct and linearly dequantize other data as it is.

[0015]

According to the image coding method of the present invention, when the image coding apparatus performs coding based on the input image data, first, the brightness correction means performs the brightness correction according to the output characteristic of the output device. That is, the frequency characteristic and / or the γ characteristic is corrected, the luminance data having a value substantially proportional to the luminance actually output by the output device is generated, and the actual output state of the output device is simulated. The generated luminance data is subjected to orthogonal transformation or non-orthogonal transformation, quantization, variable length coding, etc. to generate coded data, which is sent to a transmission line or a storage device.

In this series of processes, a quantization error occurs during quantization as in the conventional case. The quantization error is
It is added in the state where γ-correction is canceled, and it is displayed as noise in the form of conversion base in the coordinate space of the value that is almost proportional to the luminance output by the output device during decoding processing, so it has low sensitivity and high sensitivity. The generation of a DC component having a high sensitivity to the quantization noise of the bandpass filter and a low-frequency noise signal is suppressed.

[0017]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block configuration diagram of an image coding system according to an embodiment of the present invention, in which an image is transmitted through an image coding device 1 and a transmission line or storage device (hereinafter referred to as a transmission line in this embodiment) 32. It is composed of the decoding device 2.

As in the conventional example, the camera 31 is used as the image input device and the display 34 is used as the output device. The image encoding device 1 includes a frequency characteristic correction unit 11, an inverse γ correction unit 12, a conversion unit 13, and a quantization unit 1.
4, a variable length coding unit 15, and a code generation unit 16,
The image decoding device 2 includes a code decomposition unit 21, a variable length decoding unit 22, an inverse quantization unit 23, an inverse transformation unit 24, a γ correction unit 25,
And a frequency characteristic inverse correction unit 26.

In the image coding apparatus 1, the frequency characteristic correction unit 11 performs correction for simulating the frequency characteristic of the display 34 on the image data input from the camera 31. The inverse γ correction unit 12 performs inverse γ correction that simulates the γ characteristic on the image data whose frequency characteristic has been corrected, and obtains numerical data that is substantially proportional to the brightness displayed on the display 34. The conversion unit 13 performs conversion on this numerical data to obtain conversion coefficients. The quantizing unit 14 quantizes the transform coefficient input from the transforming unit 13 by a quantizing step obtained from the coefficient position. The variable length coding unit 15 variable length codes the transform coefficient quantized by the quantizing unit 14. The code generation unit 16 combines the variable-length code obtained by the variable-length coding unit 15 and header information (display control information) such as the size and gradation of the input image into encoded data.

Further, in the image decoding apparatus 2, the code decomposition unit 21 receives the above-mentioned coded data from the transmission line 32 and decomposes this into header information and variable length code. The variable length decoding unit 22 performs variable length decoding of the variable length code obtained by the code decomposition unit 21 by performing a process in the reverse procedure of the variable length coding unit 15. The inverse quantization unit 23 uses the variable length decoding unit 2
The numerical value obtained in 2 is subjected to the inverse procedure of the quantizing unit 14 and inversely quantized to obtain the transform coefficient. Inverse conversion unit 24
Performs the reverse procedure of the conversion unit 13 on the transform coefficient obtained by the inverse quantization unit 23. The γ correction unit 25 performs the reverse procedure of the inverse γ correction unit 12 on the data calculated by the inverse conversion unit 24 to perform γ correction. The frequency characteristic inverse correction unit 26 performs the reverse procedure of the frequency characteristic correction unit 11 on the data calculated by the γ correction unit 25 to obtain a decoded image.

Next, the operation of each section of the image coding apparatus 1 when image data is input will be specifically described.

The image data to be encoded is input to the frequency characteristic correction unit 11 for each pixel and subjected to frequency characteristic correction. Here, the image data to be encoded is digital data and has a natural number gradation, for example, 256 gradations, and the image data subjected to the frequency characteristic correction has a real number value.

The frequency characteristic correction output is displayed on the display 34.
When a CRT (cathode ray tube) is used, the scanning is performed in the directions shown in the order (1) to (3) as shown in FIG.

That is, considering the scanning direction of the CRT, the image into which the pixel of interest is input is X _i , and the frequency characteristic correction output is Y _i.
Assuming that the input image data of the pixel scanned immediately before the pixel of interest and the frequency characteristic correction output are X _i-1 and Y _i-1 , respectively, the frequency characteristic correction output Y _i is calculated by the following equation.

[0025]

## EQU1 ## Y _i = aX _i + bX _i-1 where a and b are different parameters depending on the output device, but in the case of a normal CRT, a is 5/6 and b is 1
The value may be about / 6.

On the other hand, the inverse γ correction section 12 obtains the inverse γ correction output Z from the frequency characteristic correction output Y (Y _i ) for each pixel by the following equation.

[Equation 2] The inverse γ-correction output thus obtained has a value that is substantially proportional to the brightness actually displayed on the display 34 in the target pixel.

The conversion unit 13 performs conversion processing such as orthogonal conversion on the entire image data obtained by the inverse γ correction unit 12. In the still image encoding method such as JPEG method,
The DCT for every 8 × 8 pixels is generally the motion compensation prediction + DCT for every 8 × 8 pixels in a moving image coding method such as the MPEG-VIDEO method. Where 8x
Orthogonal transforms such as orthonormal Wavelet transform and Hadamard transform instead of DCT for every 8 pixels, and bi-orthogonal Wavel
It is also possible to configure the image coding method using non-orthogonal transformation such as et transformation.

The quantizing unit 14 quantizes the transform coefficient obtained in the transforming unit 13 with an appropriate step width in accordance with the transform coefficient position. At this time, among the transform coefficients obtained from the transform unit 13, the coefficient corresponding to the DC component is linearly quantized after γ correction, and the other coefficients are linearly quantized as they are. By this quantization, the transform coefficient, which was previously expressed as a real number, is rounded to an integer value.
Quantization error occurs. In the variable length coding unit 15, the transform coefficient quantized in the quantization unit 14 is variable length coded to be a variable length code.

After all the variable length codes of the input image have been created by the above procedure, the code generation unit 16 calculates the variable length code obtained by the variable length coding unit 15 and the size and gradation of the image. The header information (control information required for screen display) is combined to form encoded data, which is output to the transmission path 32 or the storage device.

The image decoding apparatus 2 creates a restored image with less visual deterioration with respect to the original image by performing the reverse procedure process in the image encoding apparatus 1 as described above. The operation of each unit will be specifically described below.

The code decomposition unit 21 decomposes the encoded data input from the transmission path 32 into variable length codes and header information, and sends the header information to a screen control unit (not shown). In the variable length decoding unit 22, the variable length code obtained by the code decomposing unit 21 is subjected to the reverse procedure of the variable length coding unit 15 to perform variable length decoding. In the inverse quantization unit 23, the variable-length decoded data obtained by the variable-length decoding 22 is subjected to the inverse procedure of the quantization unit 14 to perform inverse quantization. At that time, of the variable length decoded data, the data corresponding to the DC component is subjected to linear dequantization and then inverse γ corrected,
The other data is linearly inversely quantized.

The inverse transforming section 24 subjects the data obtained by the inverse quantizing section 23 to the procedure of the inverse procedure of the transforming section 23. The inverse γ correction unit 25 also includes an inverse conversion unit 24 for each pixel.
The output Y'is obtained from the input data Z'obtained by the following equation.

[Equation 3]

In the frequency characteristic inverse correction unit 26, the data obtained from the inverse γ correction unit 25 is subjected to the processing in the reverse procedure of the frequency characteristic correction unit 11 for each pixel to obtain restored image data.
Specifically, considering the scanning direction of the display 34,
The data input from the inverse γ correction unit 25 of the target pixel is set to Y ′.
_i , the frequency characteristic inverse correction output is X ′ _i, and the input data of the pixel scanned k pixels before the target pixel and the frequency characteristic inverse correction output are Y ′ _ik and X ′ _ik , respectively. The inverse correction output _X'i is calculated.

[0034]

[Equation 4]

Here, the above equation is a series of infinite terms,
In a normal CRT, a and b are a = 5 / as described above.
Since 6, b = 1/6, a sufficient accuracy can be obtained with only some terms from the beginning, and the remaining terms can be ignored.

As described above, the image coding apparatus 1 corrects the frequency characteristic and γ characteristic of the output device before performing the conversion process on the image data, and obtains a numerical value proportional to the brightness displayed by the output device. Add correction so that
The corrected value is converted, quantized, and variable-length coded to obtain variable-length coded data. In this series of processing steps, a quantization error occurs during quantization, and this quantization error is in the form of the transformation basis in the coordinate space of the numerical value proportional to the brightness displayed by the output device during the decoding process. Since it is displayed as noise of a high-pass filter, a DC signal and a low-frequency noise signal with high sensitivity do not occur with respect to the quantization noise of a low-pass filter with low sensitivity. The noise can be effectively removed.

On the other hand, since the image decoding apparatus 2 performs the reverse processing of the encoding processing by the image encoding apparatus 1, it is possible to obtain a restored image with little visual deterioration with respect to the original image data.

The preferred embodiment of the present invention has been described above, but the present invention is not limited to this embodiment, and can be implemented in various other different modes without departing from the scope of the invention. It is possible to For example, one of the frequency characteristic correction unit 11 and the inverse γ correction unit 12 of the image coding apparatus 1 may be removed according to the types or characteristics of the input device and the output device, and the image decoding apparatus correspondingly may be removed. One of the second γ correction unit 25 and the frequency characteristic inverse correction unit 26 may be removed.

[0039]

As is apparent from the above description, in the image coding method of the present invention, the frequency characteristic and / or the γ characteristic of the image data is changed according to the characteristic of the output device before the conversion processing is performed on the image data. It is a configuration that performs correction, and the quantization error that occurs during quantization is displayed as noise in the form of a transform base in a numerical coordinate space that is approximately proportional to the brightness displayed by the output device during the decoding process. Therefore, there is an effect that generation of a DC component having high sensitivity and low-frequency noise with respect to the quantization noise of the high-pass filter having low sensitivity is significantly suppressed as compared with the related art.

Further, when the brightness correction means includes an inverse γ correction section, among the conversion coefficients obtained by the conversion means, the coefficient corresponding to the DC component is γ-corrected and then linearly quantized, The coefficient is linearly quantized as it is, and at the time of the decoding process, the data corresponding to the DC component of the variable-length decoded data is inversely γ-corrected after the linear dequantization, and the other data is linearly inversely quantized. Since it is quantized, the noise of the DC component caused by the quantization error of the DC component is less visually stimulated while suppressing the generation of the DC component and the low frequency noise caused by the quantization error of the high frequency component. effective.

As a result, it is possible to efficiently perform coding with less visual deterioration, and to realize an image coding / decoding process that can improve the compression efficiency and the quality of the restored image as compared with the prior art.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of an image coding system according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing that processing is performed according to the scanning direction of an output device.

FIG. 3 is a block configuration diagram of a conventional representative image encoding system.

4A is an explanatory diagram of a γ characteristic of an output device such as a CRT, and FIG. 4B is an explanatory diagram showing an input / output characteristic of a γ correction performed to cancel the γ characteristic of an output device such as a CRT. Fig.

[Explanation of symbols]

 1 Image Coding Device 11 Frequency Characteristic Correction Unit 12 Inverse γ Correction Unit 13 Transforming Unit 14 Quantizing Unit 15 Variable Length Encoding Unit 16 Code Generating Unit 2 Image Decoding Device 21 Code Decomposing Unit 22 Variable Length Decoding Unit 23 Inverse Quantization Conversion unit 24 inverse conversion unit 25 γ correction unit 26 frequency characteristic inverse correction unit

Continuation of front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display area H04N 5/202

Claims (5)

[Claims] 1. An image encoding device for inputting image data for displaying an image by outputting to an output device to generate predetermined encoded data, and an image encoding device for performing an inverse generation process of the encoding. In an image coding method, comprising: an image decoding device that decodes encoded data into an original image, and outputs the decoded data to the output device, wherein the image encoding device performs inverse luminance correction on input image data. And a brightness correction means for generating brightness data simulating the brightness output by the output device, and a conversion means for performing a orthogonal transformation or a non-orthogonal transformation on the generated luminance data to obtain a transformation coefficient. A quantizing means for quantizing the transform coefficient, variable length coding means for performing variable length coding on the quantized transform coefficient, and variable length coding means for generating a variable length code, An image coding method, comprising: a code generation unit that adds predetermined control information to the generated variable-length code to generate coded data. 2. The image coding method according to claim 1, wherein the brightness correction means applies an inverse γ to the input image data.
An image coding system comprising at least one of an inverse γ correction unit for performing correction and a frequency characteristic correction unit for performing luminance correction according to the frequency characteristic of the output device. 3. The image coding method according to claim 1, wherein the quantizing unit includes a transform coefficient obtained from the transforming unit when the brightness correcting unit includes the inverse γ correcting unit. An image coding method characterized in that the coefficient corresponding to the DC component is linearly quantized after γ correction, and the other coefficients are linearly quantized as they are. 4. The image coding method according to claim 1, wherein the image decoding device is provided in the code generation means for the coded data sent from the image coding device. Code decomposing means for performing reverse procedure processing to decompose into a variable length code and the control information, and reverse procedure processing of the variable length coding means for the variable length code to generate variable length decoded data. Variable-length decoding means; dequantizing means for performing inverse procedure processing of the quantizing means on the variable-length decoded data to generate dequantized data; and transforming the dequantized data. Image encoding, which comprises an inverse transforming means for performing the inverse procedure processing of the means, and an inverse luminance correcting means for performing the inverse procedure processing of the luminance correcting means on the data obtained by the inverse transforming means. method. 5. The image coding method according to claim 4, wherein the dequantization unit includes a DC component of the variable length decoded data when the luminance correction unit includes the inverse γ correction unit. An image coding method characterized in that the corresponding data is subjected to linear dequantization and then subjected to inverse γ correction, and the other data is subjected to linear dequantization as it is.