JPH11252551A - Optimizing method and system for quantization table for image coding, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an optimizing method and its system for a quantization table in image coding for optimizing quantization at a high compression rate with excellent image quality and to obtain a recording medium to record its processing program. SOLUTION: A coding section 3 uses a quantization table to encode compressed image data to be coded into coded image data, and a decoding section 4 uses the quantization table to decode the coded image data into decoded image data. Then an image compression rate calculation section 8 calculates a compression rate for the compressed image data and the coded image data, and an image quality evaluation section 9 calculates an image quality evaluation scale resulting from evaluating the image quality of the decoded image data in comparison with that of the compressed image data. A controller 21 of a high dimension algorithm arithmetic section 1 updates the quantization table based on the value of the an evaluation function based on the arithmetic value and optimizes the quantization table by repeating the processing of each section above for a prescribed number of times.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of optimizing a quantization table used in encoding and decoding of image data processed in a digital camera, a digital video camera, a digital VTR, a digital copying machine, a printer and the like. The present invention relates to an apparatus and a recording medium on which a processing program of the above-described optimization method is recorded.

[0002]

2. Description of the Related Art JP, which is one of the conventional image coding systems,
As optimization of the quantization table of the Joint Photographic Experts Group (EG) image coding method, a method based on a genetic algorithm or discrete cosine transform (Descrete Cosin
e Transform: Hereinafter referred to as DCT. A method based on the statistical properties of the coefficient of (1) has been proposed.

Hiramatsu et al. Have proposed a method for optimizing a quantization table based on a genetic algorithm (for example, see the prior art document "Harimatsu Hiramatsu et al.," Study on high-performance image compression by genetic algorithm ", Proceedings of the Society of System Control and Information Engineers Conference, Vol.38th,
pp. 37-38, 1994. " ). This method is a method for optimizing a quantization table for JPEG encoded image data by applying a genetic algorithm by regarding the quantization table as a gene.
It is a feature of this optimization method that subjective judgment is adopted in so-called selection in a genetic algorithm. In this optimization method, each value of a quantization table corresponding to a gene is set to 448 bits (64 × 7 bits).
And selects, by subjective judgment, a quantization table emphasizing image quality as an initial individual from among quantization tables generated by random numbers. The number of individuals and the number of generations are 6
Individuals and 100 generations are set, and in each generation, 3 individuals are selected from the above 6 individuals and used as parents of the next generation. 1 for crossover
Using point crossover, mutation is a bit inversion with a 1% probability per bit. It is reported that a quantization table with the best image quality was obtained in the 80th generation. When encoding is performed using the quantization table, the compression ratio is higher than when encoding is performed using the first-generation quantization table. However, it is also reported that the compression ratio does not change as compared with a quantization table of an example value for JPEG encoding.

Have proposed a method of optimizing a quantization table based on the statistical properties of DCT coefficients (for example, see the prior art document "T. Eude et a").
l., “Statistical distribution of DCT coefficients
and their application to anadaptive compression al
gorithm ”, Tencon, Vol. 1, pp. 427-430, 1994”. ).
Ude et al. Use a quantization table derived from a distribution model of DCT coefficients of a training image set to use JPE.
To obtain a better image quality with no block effect even at the same compression rate, when encoding using the G encoding method, as compared with encoding using the quantization table of the exemplary value for JPEG encoding Report that you can. The image targeted by this optimization method is a medical image, and the distribution of DCT coefficients of the medical image to be encoded is 3
It has been reported that it can be expressed by combining up to Gaussian distributions. In this optimization method, a quantization table in which the number of zeros after quantization is controlled by the distribution model is obtained. Here, the compression ratio is evaluated only by the number of zeros of the quantized DCT coefficient. In general, it is expected that the larger the number of zeros of the quantized DCT coefficient, the higher the compression ratio. When the compression ratio is fixed, the JPEG code is calculated using the quantization table obtained by this method, as compared with the case where encoding is performed by the JPEG coding method using the example value quantization table for JPEG coding. It has been reported that an image having a high SNR (signal-to-noise ratio for an image) can be obtained when encoding is performed by a coding scheme. It is also reported that this result does not change for different types of medical images such as CT (Computed Tomography) and MRI (Magnetic Resonance Imaging).

[0005]

A disadvantage of the above-described method of optimizing a quantization table by a genetic algorithm is that a quantization table that satisfies both a higher compression rate and better image quality cannot be obtained at the same time. It is. On the other hand, DCT
The disadvantage of the quantization table optimization method based on coefficient statistics is that when trying to obtain an optimized quantization table for an image, many images are required to perform statistics as a prerequisite. And that only an optimized quantization table can be obtained for an image having the same statistical properties, that is, a similar image (a medical image in the report of Ude et al.).

Therefore, there is a need for a quantization table optimizing method and apparatus which simultaneously satisfies a high compression rate and good image quality for general image compression coding.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above problems, and to provide a quantization table optimizing method and apparatus capable of encoding a general image with a high compression rate and excellent image quality. Another object of the present invention is to provide a recording medium recording a processing program of an optimization method.

[0008]

According to a first aspect of the present invention, there is provided a method for optimizing a quantization table for image encoding, wherein compressed image data to be encoded is encoded using a quantization table. An encoding step of encoding data; a decoding step of decoding the encoded image data into decoded image data using the quantization table; and a data of the compressed image data and the encoded image data. Calculates a compression ratio, which is a ratio of the amount or a bit rate (amount of information expressed in bits), which is the number of bits per pixel of the decoded image data, and compares the calculated compression ratio with the compressed image data. A calculating step of calculating an image quality evaluation scale for evaluating the image quality of the data; calculating an evaluation function value based on the compression ratio and the image quality evaluation scale; Optimizing the quantization table by repeating a predetermined number of times the update step of updating the quantization table, the encoding step, the decoding step, the calculation step, and the update step. And a controlling step.

According to a second aspect of the present invention, there is provided a method for optimizing a quantization table for image encoding, wherein the image quality evaluation scale is set as follows. , Defined by Japanese Industrial Standards Z8730, and is characterized by a color difference of an object color indicating a distance between the object color of the image of the compressed image data and the object color of the image of the decoded image data.

Further, the method for optimizing a quantization table for image encoding according to claim 3 is a method for optimizing a quantization table for image encoding according to claim 1, wherein:
The image quality evaluation scale is a signal-to-noise ratio when the maximum value of the compressed image data is a signal, and the difference between the compressed image data and the decoded image data is noise.

According to a fourth aspect of the present invention, there is provided a quantization table optimizing apparatus for image encoding according to the first, second, or third aspect. The evaluation function is a value obtained by multiplying the compression ratio by a predetermined first weighting factor and the image quality evaluation scale by a predetermined second weighting coefficient.
The update step is characterized by changing the value of the quantization table so as to increase the absolute value of the evaluation function. Here, when the evaluation function is constituted by the compression ratio and the color difference between the object colors, the first weighting coefficient and the second weighting coefficient are different codes. Further, when the evaluation function is composed of a compression ratio and a signal-to-noise ratio, the first weighting coefficient and the second weighting coefficient have the same sign.

Further, the method of optimizing a quantization table for image coding according to claim 7 is a method of optimizing a quantization table for image coding according to one of claims 1 to 6. In the method, the updating step includes adding a predetermined small value to the value of the evaluation function calculated using the current quantization table used for the encoding and decoding and the value of the current quantization table. The current value of the quantization table is updated based on the amount of change between the value of the evaluation function calculated using the quantization table and the value of the current quantization table.

According to a eighth aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding, comprising the steps of encoding compressed image data to be encoded into encoded image data using the quantization table. Decoding means, decoding means for decoding the encoded image data into decoded image data using the quantization table, and a data amount ratio or decoding between the compressed image data and the encoded image data. Calculating means for calculating a compression rate which is a bit rate which is the number of bits per pixel of the decoded image data, and calculating an image quality evaluation scale for evaluating the image quality of the decoded image data by comparing with the compressed image data. Updating means for calculating a value of an evaluation function based on the compression ratio and the image quality evaluation scale, and updating the quantization table based on the value of the evaluation function;
Control means for optimizing the quantization table by repeating each processing of the encoding means, the decoding means, the calculating means, and the updating means a predetermined number of times. And

According to a ninth aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding, wherein the image quality evaluation scale is set as follows. , Defined by Japanese Industrial Standards Z8730, and is a color difference of an object color indicating a distance between the object color of the image of the compressed image data and the object color of the image of the decoded image data.

Further, according to a tenth aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding according to the eighth aspect of the present invention. And a signal-to-noise ratio when the maximum value of the compressed image data is defined as a signal and the difference between the compressed image data and the decoded image data is defined as noise.

According to the eleventh aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding.
0, wherein the evaluation function sets the compression ratio to a predetermined first value.
And a value obtained by multiplying the image quality evaluation scale by a predetermined second weighting factor, and the updating means increases the absolute value of the evaluation function. The value of the quantization table is changed. Here, when the evaluation function is constituted by the compression ratio and the color difference between the object colors, the first weighting coefficient and the second weighting coefficient are different codes. Further, when the evaluation function is composed of a compression ratio and a signal-to-noise ratio, the first weighting coefficient and the second weighting coefficient have the same sign.

Further, the apparatus for optimizing a quantization table for image encoding according to claim 14 is provided.
In the apparatus for optimizing a quantization table for image encoding according to any one of the above, the updating unit performs an operation by the arithmetic unit using a current quantization table used for the encoding and decoding. Between the calculated value of the evaluation function and the value of the evaluation function calculated by the calculation means using the quantization table when a predetermined minute value is added to the value of the current quantization table. , The current value of the quantization table is updated.

According to a fifteenth aspect of the present invention, there is provided a recording medium according to any one of the first to seventh aspects, wherein each of the steps is included in a method for optimizing a quantization table for image encoding. Is recorded.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an example of an image processing system for implementing a quantization table optimizing method according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a configuration example of an image processing system according to an embodiment of the present invention. The image processing system according to the present embodiment includes (a) a high-dimensional algorithm operation unit 1 that implements an algorithm for optimizing a value of a quantization table, and (b) a high-dimensional algorithm operation unit 1
A quantization table storage unit 2 for storing a value of a quantization table for encoding and decoding or a value of a temporary quantization table in the course of optimization optimized by (c).
JPEG encoding of the compressed image data to be compressed stored in the compressed image storage unit 5 using the value of the quantization table stored in the quantization table storage unit 2
JPEG decoding in which the encoding unit 3 and (d) JPEG decoding of the JPEG image data to be decoded stored in the JPEG image storage unit 6 using the values of the quantization table stored in the quantization table storage unit 2 Unit 4, (e) a compressed image storage unit 5 for storing compressed image data to be compressed, which is a target of image processing, and (f) JPEG encoded image data compressed by the JPEG encoding unit 3. JPEG image storage unit 6, (g) decoded image storage unit 7 for storing the decoded image data decoded by JPEG decoding unit 4, and (h) compressed image storage unit 5 By using the compressed image data and the JPEG encoded image data stored in the JPEG image storage unit 6, the ratio of the size of each image data in byte units is calculated. And (i) an image for evaluating the quality of the compressed image stored in the compressed image storage unit 5 and the quality of the decoded image stored in the decoded image storage unit 7 And a quality evaluation unit 9.

The high-dimensional algorithm operation unit 1 implements an algorithm for optimizing the value of the quantization table, and the image compression ratio and the image quality evaluation unit 9 which is the output value of the image compression ratio calculation unit 8. The value of the quantization table is optimized using the image quality evaluation value that is the output value, and the result value is output to the quantization table storage unit 2. As shown in FIGS. 2 and 3, the high-dimensional algorithm operation unit 1
1, a coordinate initial value storage device 22, and a coordinate storage device 23
, A coordinate quantization step conversion device 24, a quantization table temporary storage device 25, and a quantization table reading device 2
6, a coordinate calculation device 27, and a coordinate storage device 28 before update
, Coordinate difference calculation device 29, coordinate difference storage device 210
, A minute time constant storage device 211, and a speed calculation device 21
2, the speed storage device 213, and the minute coordinate difference calculation device 2
14, an attenuation constant storage device 215, a minute coordinate difference storage device 216, a compression ratio storage device 217, an evaluation function operation device 218, a pre-update evaluation function value storage device 219, a force operation device 220, an image quality An evaluation value storage device 221;
Evaluation function value storage device 222 after update and force storage device 223
, The temporary coordinate storage device 224 and the database device 22
5 is provided.

The quantization table storage unit 2 is a storage device for storing the values of the quantization table optimized by the high-dimensional algorithm operation unit 1 or the values of the temporary quantization table during the optimization. . Quantization table storage unit 2
Is composed of a general memory.

The JPEG encoding unit 3 reads out the compressed image data to be compressed stored in the compressed image storage unit 5 and stores the compressed image data in the quantization table stored in the quantization table storage unit 2. JPEG encoding is performed using the values, and the resulting value, JPEG encoded image data (hereinafter, referred to as encoded image data), is output to the JPEG image storage unit 6. As shown in FIG. 4, the JPEG encoding unit 3 reads the compressed image data to be compressed stored in the compressed image storage unit 5 and outputs the compressed image data to the color conversion device 32, The image processing apparatus includes a color conversion device 32, a pixel thinning device 33, a DCT operation device 34, a quantization device 35, a Huffman coding device 36, and a JPEG image reading device 37. The JPEG encoding unit 3 reads an image to be encoded and converts it into image data by a known method, and then executes a color conversion process, a pixel thinning process, a DCT operation process, a quantization process, and a Huffman encoding process. JPEG encoding, and the encoded image data
Output from the PEG image reading device 37.

The JPEG decoding unit 4 reads out the JPEG image data to be decoded stored in the JPEG image storage unit 6 and uses the JPEG image data by using the value of the quantization table stored in the quantization table storage unit 2. J
PEG decoding is performed, and the result value is output to the decoded image storage unit 7. As shown in FIG. 5, the JPEG decoding unit 4 includes a JPEG image reading device 41, a Huffman decoding device 42, an inverse quantization device 43, and an IDCT operation device 4.
4, a pixel interpolating device 45, a color converting device 46, and a decoded image reading device 47. JPEG decoding unit 4
In a known method, Huffman decoding processing, inverse quantization processing, IDCT operation processing which is an inverse discrete cosine transform (IDCT), pixel interpolation processing, color conversion processing, Is performed, the decoded image data (hereinafter, referred to as decoded image data) subjected to JPEG decoding is output from the decoded image reading device 47.

The compressed image storage section 5 stores image data to be compressed which is to be subjected to image processing. The JPEG image storage unit 6 stores the encoded image data compressed and encoded by the JPEG encoding unit 3. The compressed encoded image data is input to the JPEG decoding unit 4. The decoded image storage unit 7 stores the decoded image data decoded by the JPEG decoding unit 4. here,
The compressed image storage unit 5, the JPEG image storage unit 6, and the decoded image storage unit 7 are configured by a general memory.

The image compression ratio calculation section 8 reads out the compressed image data stored in the compressed image storage section 5 and the encoded image data stored in the JPEG image storage section 6 and determines the size of each image data. By calculating the ratio of the image data in units of bytes, the compression ratio of the image data is calculated, and the operation result value is output to the high-dimensional algorithm operation unit 1. For example, the compression ratio of image data is defined as a value obtained by dividing the data size of compressed image data by the data size of encoded image data.

The image quality evaluator 9 includes a compressed image storage 5
, And the decoded image data stored in the decoded image storage unit 7 are read, the image quality evaluation values are calculated, and the calculation result value is sent to the high-dimensional algorithm calculation unit 1. Output. The image quality evaluation value, which is an image quality evaluation scale, is preferably a Japanese Industrial Standard (JIS) Z873 as an image quality evaluation scale.
0, which is the distance between the object color of the compressed image and the object color of the decoded image (hereinafter referred to as the color difference of the object color), or the difference between the compressed image and the decoded image. A signal-to-noise ratio (hereinafter, referred to as SNR) or the like is used. These values are calculated as follows.

The color difference between the object colors is determined by the Japanese Industrial Standard (JI
S) According to the definition in Z8730, it is the color difference of the object color measured in the L * a * b color system (CIELAB color system). If the color difference of the object color is ΔE, the color difference ΔE of the object color is It is calculated using the following equation (1).

ΔE = √ (ΔL * ² + Δa * ² + Δb * ² ) where ΔL * is the CIE1976 lightness L * of the object color of the image of the compressed image data and the object color of the image of the decoded image data. Δa * is a color coordinate a * between the object color of the image of the compressed image data and the object color of the image of the decoded image data.
Δb * is the difference between the color coordinates b * of the object color of the image of the compressed image data and the object color of the image of the decoded image data. The color difference for each pixel is obtained by using the above equation 1, and the average color difference obtained by averaging the color differences for the entire image is used as the image quality evaluation value. Hereinafter, the average color difference is simply referred to as color difference. The “color difference” of the color difference signal, which is the color component of the image,
Is different from

For the SNR, the maximum value of the pixel values of the compressed image data is used as a signal. For example, if the image data is 8 bits / pixel, 255 is used as a signal, and the decoded image data and noise are used as noise. This is a signal-to-noise ratio value using a mean square error n with the compressed image data, and is calculated using the following equation (2).

SNR = 20 · log {255 / {(n)} where the mean square error n is a value calculated by one of the following equations (a) to (c). . Hereinafter, M and N are the number of pixels of the width and height of the image, respectively, and (i, j) is the position (coordinate value) of the pixel.

(A) When only luminance component is considered

[0031]

(Equation 3)

Here, Y _O (i, j) is a luminance component of the compressed image data, and Y _D (i, j) is a luminance component of the decoded image data.

(B) Considering all color components (RGB color system)

[0034]

Equation 4] M N n = {1 / ( 3MN)} × [Σ Σ {R O (i, j) -R D (i, j)} 2 i j M N + Σ Σ {G O (i, j _{) -G D (i, j)} } 2 i j M N + Σ Σ {B O (i, j) -G D (i, j)} 2] i j

Here, the RGB color system is a three-color system based on the CIE color system, and is a color system based on three monochromatic radiations of red R, green G and blue B. is there. Also, R
_O (i, j) and R _D (i, j) are red components (R components) of the compressed image data and the decoded image data, respectively, and G _O (i, j) and G _D (i, j). ) Are the green components (G components) of the compressed image data and the decoded image data, respectively, and B _O (i, j) and G _D (i, j) are the compressed image data and the decoded image data, respectively. This is the blue component (B component) of the data.

(C) When all color components are considered (when represented by luminance Y and color differences Cr and Cb)

[0037]

(Equation 5)

Here, Y _O (i, j) and Y _D (i, j)
Are the Y components (luminance components) of the compressed image data and the decoded image data, respectively, Cb _O (i, j) and Cb _D
(I, j) are Cb components (color difference Cb components) of the compressed image data and the decoded image data, respectively, and Cr
_O (i, j) and Cr _D (i, j) are Cr components (chrominance Cr components) of the compressed image data and the decoded image data, respectively.

Next, the control device 21 will be described in detail.
The control device 21 manages and controls the processing of the high-dimensional algorithm operation unit 1. The control device 21 includes, as shown in FIG. 6, an arithmetic circuit 51, a control circuit 52, and a storage circuit 53.
A minimum coordinate value holding register 54, a maximum coordinate value holding register 55, a time counter 56, a maximum time holding register 57, a frequency counter 58, a maximum frequency holding register 59, a speed absorption constant holding register 510, An internal bus 511 is provided.

The operation circuit 51 performs various operations of the control device 21 and the entire system. For example, a comparison operation for determining the end of the repetitive processing using the time counter 56, the maximum time holding register 57, the frequency counter 58, and the maximum frequency holding register 59, which will be described later, is performed. The control circuit 52 performs various management controls of the control device 21 itself and the entire system. The storage circuit 53 stores a processing program necessary for management, control, or operation of the control device 21 and the entire system, and data necessary for executing the processing program.

The minimum coordinate value holding register 54 holds the minimum coordinate value q _min which is the lower limit value of the coordinate value, and the maximum coordinate value holding register 55 stores the maximum coordinate value q which is the upper limit value of the coordinate value.
Keep _max . Further, the time counter 56 holds a current value t of the number of repetitions. Here, the number of repetitions or the number of repetitions in a first repetition control process to be described later is “time”.
That. The maximum time holding register 57 holds a maximum time t _max which is the maximum value of the number of repetitions (time). Further, the frequency counter 58 holds the current frequency value i in the DCT operation. DCT of JPEG encoder 3
The arithmetic unit 34 and the IDCT arithmetic unit 44 of the JPEG decoding unit 4 calculate the DCT coefficient corresponding to the frequency i held in the frequency counter 58.

In the image processing system according to the embodiment of the present invention, the values, coordinates, pre-update coordinates, and speed of the quantization table are calculated for each frequency. The maximum frequency holding register 59 holds the maximum value i _max of the frequency i. Here, when optimizing only the quantization table for the luminance signal or only the quantization table for the chrominance signal, the maximum frequency i _max to be held by the maximum frequency holding register 59 is 64. When the table and the color difference signal quantization table are simultaneously optimized, the maximum frequency i _max to be held by the maximum frequency holding register 59 is 128. Therefore, when optimizing only the luminance signal quantization table or only the color difference signal quantization table, the frequency i indicated by the frequency counter 58 is from 1 to 64, and the luminance signal quantization table and the color difference signal When optimizing the quantization table at the same time, the frequency counter 5
The frequency i indicated by 8 is from 1 to 128. In addition,
When the quantization table for the luminance signal and the quantization table for the chrominance signal are simultaneously optimized, and when the value i indicated by the frequency counter 58 indicates 1 to 64, processing for each frequency of the quantization table for the luminance signal is performed. When the frequency value i indicated by the frequency counter 58 indicates 65 to 128, processing for each frequency of the color difference signal quantization table is performed.

The speed absorption constant holding register 510 holds the speed absorption constant absorption used in the reflection processing (step S85) described later. Furthermore, internal bus 5
11 is an arithmetic circuit 51, a control circuit 52, a storage circuit 5
3, a minimum coordinate value holding register 54, a maximum coordinate value holding register 55, a time counter 56, a maximum time holding register 57, a frequency counter 58, a maximum frequency holding register 59, and a speed absorption constant holding register 510. Are connected to each other.

Next, the quantization table optimization process stored in the quantization table storage unit 2 of FIG. 1 and executed by the high-dimensional algorithm operation unit 1 will be described with reference to the flowchart of FIG. This quantization table optimization process includes an initialization process S11, a step S12 for determining the end,
Coordinate quantization step conversion processing S13, quantization table reading processing S14, image evaluation processing S15, evaluation function calculation processing S16, partial updating processing S17, updating processing S18, and processing S for incrementing the time counter 56.
19, a database registration process S20, and a database search process S21.

The initialization process S11 is an initialization process for optimizing the quantization table stored in the quantization table storage unit 2, and performs a process such as setting the initial value of the time t to 1. . Optimization of the quantization table includes coordinate quantization step conversion processing S13, quantization table reading processing S14, image evaluation processing S15, evaluation function calculation processing S16, partial update processing S17, and update processing S18.
And step S19 of incrementing the time counter 56 are repeated until the end condition of step 12 is satisfied.

In step S12 for determining the end, referring to FIG. 6, the arithmetic circuit 51 determines the value of the time t held in the time counter 56 and the value of the maximum time t _max held in the maximum time holding register 57. If the end condition, Equation 6 of the following equation, is not satisfied, then the coordinate quantization step conversion processing S13 is executed, and if satisfied, the iterative processing is terminated and the quantization table optimization processing is performed. finish.

[0047]

T> t _max Therefore, each of the processes S13 to S19 in the iterative process is t
Executed _max times.

The coordinate quantization step conversion process S13 is a process of associating a quantization step, which is a value of a quantization table, with a corresponding coordinate. In a general JPEG encoding device or JPEG decoding device, a quantization step which is a non-negative integer value is used as a value of a quantization table. In the quantization table optimization process according to the embodiment of the present invention, optimization is performed using coordinates corresponding to a quantization step, and the coordinates are 0.0, which is a value obtained by dividing the quantization step by 255.0. A real number normalized to a value from to 1.0 is used. Therefore, it is necessary to convert the coordinate values into a quantization step. That is, the normal quantization step has an integer value from 1 to 255. 255 is the maximum value of a general quantization step. In the coordinate quantization step conversion processing S13 of the present embodiment, the coordinate quantization step conversion device 24 sets the value of the coordinate q ₁ (i) (where i represents the frequency) stored in the coordinate storage device 23. And calculate the product of those values and 255.0,
The result value is output to the quantization table temporary storage device 25. The result value is a real value, but the real value may be converted to an integer to generate a quantization table so as to be applicable to a general JPEG encoding device and JPEG decoding device. This coordinate quantization step conversion processing S13 is performed on the coordinates corresponding to all the frequencies.

In the quantization table reading process S14, the value of the quantization table is read from the high-dimensional algorithm operation unit 1 in FIG. Specifically, referring to FIG. 2, the quantization table reading device 26 reads out each value of the quantization table stored in the quantization table temporary storage 25, that is, each quantization step, and illustrates the quantization step. 1 is output to the quantization table storage unit 2 and stored.

In the image evaluation processing S15, the compressed image data is JPEG-encoded into JPEG-encoded image data using the quantization table read out to the quantization table storage unit 2 by the quantization table reading processing S14. after JPEG decoding the JPEG encoded picture data to the decoded image data, further, the compression rate storage unit the compression ratio C _1o calculates the compression ratio C _1o based on the compressed image data and the coded image data 217 and stores the image quality evaluation value C _2o indicating the quality of the image based on the compressed image data and the decoded image data, and stores the image quality evaluation value C _2o in the image quality evaluation value storage device. 2
21 for storage.

In the evaluation function calculation processing S16, the pre-update evaluation function value U _old is calculated using the following equation based on the compression ratio C _1o and the image quality evaluation value C _2o calculated in the image evaluation processing S15. Specifically, referring to FIG. 3, the evaluation function calculation device 218 reads out the compression ratio C _1o from the compression ratio storage device 217 and the image quality evaluation value C _2o from the image quality evaluation value storage device 221, The pre-update evaluation function value U _old is calculated based on the result, and the result value is output to the pre-update evaluation function storage device 219 and stored.

Here, the evaluation function U for calculating the pre-update evaluation function value U _old and the post-update evaluation function value U _new described later is defined by the following equation (7).

U = w ₁ C ₁ + w ₂ C ₂ where w ₁ and w ₂ are the first and second weighting factors, respectively. When the image quality evaluation scale is the color difference of the object color, preferably -0.01 and 0.99. C ₁ is a compression ratio, and C ₂ is an image quality evaluation value.

When the image quality evaluation scale is SNR, the first and second weighting factors w ₁ and w ₂ are preferably

## EQU8 ## w ₁ = −0.01 and w ₂ = −0.99.

Next, the partial update process S17 is a process of calculating a coordinate difference or the like for calculating a coordinate value to be updated for each frequency by a high-dimensional algorithm. The update process S18 is a process of updating the coordinate value for each frequency using the coordinate difference calculated in the partial update process S17. In the database registration process S20, the control device 21 transmits the quantization table at time t,
The compression ratio C ₁ , the image quality evaluation value C _2, and the evaluation function value are calculated at time t.
At the same time, a process of registering in the database device 225 is performed. Further, in the incrementing process S19 of the time counter 56, the value of the time t stored in the time counter 56 is incremented by 1, and the process returns to step S12. When the end condition of step S12 is satisfied, a database search process S21 is performed. The database search processing S21 is performed by the control device 21 such that the maximum compression ratio, the minimum image quality evaluation value when the image quality evaluation value is a color difference, the maximum image quality evaluation value or the maximum evaluation function value when the image quality evaluation value is a signal to noise ratio, A search is performed from the database device 225 for a desired quantization table using the as a search key. Here, it is assumed that a desired maximum compression ratio or the like serving as a search key is set in advance by the user.

Hereinafter, the initialization processing S11, the image evaluation processing S15, the partial update processing S17 and the update processing S18 will be described in detail.

As shown in FIG. 8, the initialization process S11 includes a control device initialization process S21, a coordinate storage device initialization process S22, and a minute time constant storage device initialization process S23.
And a speed storage device initialization process S24, a pre-update coordinate storage device initialization process S25, a damping constant storage device initialization process S26, and a first iterative control initialization process S27.

In the control device initialization process S21, various initializations for the control device 21 are performed. For example, each parameter is initialized as in the following equation.

[0058]

## EQU9 ## q _min = 1.0 / 255.0

## _EQU10 ## q _max = 1.0

## _EQU11 ## t _max = 500

## _EQU12 ## i _max = 128

## EQU13 ## absorption = 0.5

Here, q _min is the minimum coordinate value and is stored in the minimum coordinate value holding register 54. q _max is the maximum coordinate value and is stored in the maximum coordinate value holding register 55. t _max is the maximum time, and the maximum time holding register 5
7 is stored. i _max is the maximum frequency and is stored in the maximum frequency holding register 59. Velocity absorption constant
absorption is stored in the velocity absorption constant register 510.

In the coordinate storage device initialization process S22, the coordinate storage device 23 is initialized. For example, a value obtained by dividing a so-called example value which is a data value of the quantization table by 255.0 is represented by a coordinate q ₁ for each frequency i.
Set as the initial value of (i). Here, the example value of the quantization table refers to the related art document “ISO / IEC10918-1: 1994
(E) “Information technology-Digital compression an
d coding of continuous-tone still images: Requireme
nts and guide lines ”, Annex K, Table K.1-Luminanc
e quantization table and Table K.2-Chrominance quan
tization table ". That is, when i = 1, for example, q ₁ (1) = 16/255
Is set.

In the small time constant storage device initialization processing S23, the initialization for the small time constant storage device 211 in FIG. 3 is performed. For example, a small time constant storage d
t is set to 0.005, and the minute time constant storage device 21
1 is stored. In addition, speed storage device initialization processing S24
In, initialization of the speed storage device 213 in FIG. 3 is performed. For example, the speed v (i) corresponding to each frequency i is represented by the minimum coordinate value q _min held in the coordinate minimum value holding register 54 in FIG. 6 and the value of the minute time dt stored in the minute time constant storage device 211. Is set using the following equation (14), and the speed storage device 21
3 is stored.

[0062]

V (i) ← (q _min /dt)×0.005

In the pre-update coordinate storage device initialization processing S25, initialization is performed on the pre-update coordinate storage device 28 in FIG. For example, the coordinates q ₀ (i) before update corresponding to each frequency i are represented by coordinates q ₀ corresponding to each frequency i.
_{From 1} (i), using the speed v (i) corresponding to each frequency i and the minute time constant dt, the value is set using the following equation (15) and stored in the pre-update coordinate storage device 28.

[0064]

## EQU15 ## q ₀ (i) ← q ₁ (i) −v (i) × dt

In the attenuation constant storage device initialization processing S26, initialization is performed on the attenuation constant storage device 215 in FIG. For example, the damping constant dumping is
= 0.999, and is stored in the attenuation constant storage device 215.

In the first repetitive control initialization processing S27, the time t in the time counter 56 in FIG.
, The initialization for the first iteration control is performed. Here, the first repetitive control includes a step S12 of performing a determination process by the control device 21 using the time counter 56 and the maximum time holding register 57, a coordinate quantization step conversion process S13, and a quantization table reading process. S14, image evaluation processing S15, evaluation function calculation processing S16, partial update processing S17, and update processing S18
And control of an iterative process including the process S19 of incrementing the time counter 56.

As shown in FIG. 9, the image evaluation processing S15 includes a JPEG encoding processing S31, a JPEG decoding processing S32, a compression ratio calculation processing S33, and an image quality evaluation processing S34. Here, JPEG encoding processing S31
Is a process for compressing an image by JPEG encoding. As shown in FIG. 10, the JPEG encoding process S31 includes a compressed image reading process S41 and a color conversion process S41.
42, pixel thinning processing S43, and DCT processing S44
, Quantization processing S45, and Huffman encoding processing S46
And a JPEG image reading process S47.

In the compressed image reading process S 41, referring to FIG. 4, the compressed image reading device 31 reads the image data to be compressed from the compressed image storage unit 5 and outputs it to the color conversion device 32. In the color conversion processing S42, the color conversion device 32 converts the color space of the input compressed image data, for example, the RGB color space into the color space YCbCr of the luminance signal Y and the color difference signals Cb and Cr, and the color space is converted. The output image data is output to the pixel thinning device 33. In the pixel thinning process S43, the pixel thinning device 33 thins out the pixels of the compressed image data converted into the YCbCr color space, and outputs the result value to the DCT operation device. In the DCT processing S44, the DCT operation device 34 performs a discrete cosine transform on the compressed image data on which the pixels have been thinned out, and outputs the result value to the quantization device 35. In the quantization process S45,
The quantization device 35 reads out the quantization step at each frequency, which is the data value of the quantization table, from the quantization table storage unit 2 and quantizes the input image data using the quantization step at each frequency. The result is output to the Huffman encoding device 36.
In the Huffman encoding process S46, the Huffman encoding device 36 performs Huffman encoding on the image data quantized by the quantization device 35, and outputs a result value to the JPEG image reading device 37. In the JPEG image reading process S47, the JPEG image reading device 37
Attaches a header file for JPEG encoded image data to the above Huffman encoded image data,
It is output to the JPEG image storage unit 6 in FIG. 1 and stored as a G encoded image file.

Here, the JPEG encoding unit 3 used in the embodiment according to the present invention will be described. In general JPEG encoding, JPEG encoding is performed using a value of a quantization table that is a non-negative integer such as an example value. In the image processing system according to the embodiment of the present invention, preferably, JPEG encoding is performed using values of the quantization table, which is a non-negative real number, so as to be more suitable for the image processing system. As described above, when the value of the quantization table is a non-negative real number, the quantization process S4
5, the quantization is performed using the value of the real quantization table. However, in the JPEG image reading process S47, since the definition of the JPEG encoding method does not correspond to the value of the real quantization table, It is necessary to use a value obtained by converting the value of the real quantization table into an integer by truncation below a floating point or the like as a temporary quantization table for outputting a JPEG encoded image.

The JPEG decoding process S32 is a process for decompressing an image by the JPEG decoding method using the above quantization table. JPEG decoding processing S32
As shown in FIG. 11, the JPEG image reading process S
51, Huffman decoding processing S52, and inverse quantization processing S
53, IDCT processing S54, and pixel interpolation processing S55
, Color conversion processing S56, and JPEG image reading processing S
57.

In the JPEG image reading process S51, referring to FIG.
Is the JPEG image data to be decoded
The JPEG image data is read from the image storage unit 6 and output to the Huffman decoding device 42. In the Huffman decoding process S52, the Huffman decoding device 42 performs Huffman decoding on the input JPEG encoded image data,
The result value is output to the inverse quantization device 43. In the inverse quantization processing S53, the inverse quantization device 43 uses the value of the quantization table stored in the quantization table storage unit 2 in FIG. 1 to perform the data value Huffman decoding performed by the Huffman decoding device 42. Is inversely quantized, and the resulting value is expressed as I
Output to the DCT arithmetic unit 44. In the IDCT processing S54, the IDCT operation device 44 performs an inverse discrete cosine transform on the inversely quantized data, and outputs a result value to the pixel interpolation device 45. In the pixel interpolation process S55, the pixel interpolation device 45 performs the JPEG encoding process S3
Interpolation processing of the pixels thinned out by 1 is performed, and the result value is output to the color conversion device 46. In the color conversion processing S56, the color conversion device 46 converts the input color space of the result value from the YCbCr color space to the RGB color space, and outputs the result value to the decoded image reading device 47. In the encoded image reading process S57, the decoded image reading device 47 outputs the image data converted into the RGB color space to the decoded image storage unit 7 in FIG.

Next, the JPEG decoding section 4 will be described. In a general JPEG decoding method, a value of a quantization table which is a non-negative integer such as an example value is used to perform J
Although PEG decoding is performed, in the embodiment according to the present invention, preferably, JPEG decoding is performed using a value of a quantization table that is a non-negative real number so as to be more suitable for an image processing system. In the case where the value of the quantization table is a non-negative real number as described above, in the inverse quantization process S53, inverse quantization is performed using the value of the quantization table that is a real number, but in the JPEG image reading process S51, the inverse quantization is performed. Since the JPEG encoding method does not correspond to the value of the real quantization table, the value obtained by converting the value of the real quantization table into an integer by truncating a floating point or less is used for outputting JPEG encoded image data. Must be used as a temporary quantization table. However, the quantization table having the integer value is not used in the inverse quantization processing S53.

The compression ratio calculation processing S33 calculates the compression ratio C _1o or the compression ratio C _1o as described above. The image quality evaluation processing S34 calculates the image quality evaluation value C _2o or image quality evaluation value C _2n is a color difference or SNR of object color as described above.

Next, as shown in FIGS. 12 and 13, the partial update processing S17 includes a second iterative control initialization processing S61, an end determination step S62, and a minute coordinate difference calculation processing S63. , Coordinate value copy processing S64,
Partial coordinate value update processing S65 and coordinate range optimization processing S6
6, a coordinate quantization step conversion process S67, a quantization table reading process S68, an image evaluation process S69,
It includes an evaluation function calculation process S70, a force calculation process S71, a coordinate difference calculation process S72, a speed update process S73, and a process S74 for incrementing the frequency counter 58.

In the second iterative control initialization processing S61, referring to FIG. 6, the frequency i held in the frequency counter 58 is set to 1 to perform initialization for the second iterative control. Here, the second repetitive control is a step S62 in which the control device 21 using the frequency counter 58 and the maximum frequency holding register 59 makes an end determination.
, Small coordinate difference calculation processing S63, and coordinate value copying processing S
64, partial coordinate value update processing S65, coordinate range optimization processing S66, coordinate quantization step conversion processing S67,
Quantization table readout processing S68 and image evaluation processing S
69, evaluation function calculation processing S70, force calculation processing S71
, Coordinate difference calculation processing S72, and speed update processing S73
And the process S74 of incrementing the frequency counter 58
This is a control for an iterative process including:

In step S 62 for determining the end, the arithmetic circuit 51 compares the value of the frequency i held in the frequency counter 58 with the value of the maximum frequency i _max held in the maximum frequency holding register 59. Here, the control circuit 52 ends the iterative processing and ends the partial update processing if the end condition of the following equation 16 is satisfied, and then executes the update processing S18. On the other hand, the following expression 16
If the end condition is not satisfied, then the minute coordinate difference calculation processing S63 is executed.

[0077]

Equation 16] i> i _max Therefore, the processes in the iterative process is performed on the coordinate or the like corresponding to all frequencies.

In the minute coordinate difference calculation processing S63,
Referring to FIG. 3, the minute coordinate difference calculating device 214 stores the speed v (i) corresponding to the frequency i from the speed storage device 213.
And the value dt from the minute time constant storage device 211, calculate the minute coordinate difference dq (i) using the following equation 17, and output the result value to the minute coordinate difference storage device 216.

[0079]

## EQU17 ## dq (i) ← v (i) × dt

In the coordinate value copying process S64, referring to FIG. 2, the control device 21 determines the coordinates q corresponding to all the frequencies.
Stored in the temporary coordinate value storage device 224 as _a temporary value of (j) coordinate q ₁ p (j). Here, j is a frequency from 1 to 128. Next, the partial coordinate value update processing S65
In, the control device 21 calculates the coordinate q ₁ (i) corresponding to the frequency i from the coordinate storage device 23 and the minute coordinate difference dq corresponding to the frequency i from the minute coordinate difference storage device 216.
(I) is read out, the temporary coordinate q ₁ p (i) corresponding to the frequency i is calculated using the following equation (18), and the result value is output to the temporary coordinate storage device 224 and stored.

[0081]

## EQU18 ## q ₁ p (i) ← q ₁ (i) + dq (i)

Next, in the coordinate range optimizing process S66, the control device 21 sets the temporary coordinate q corresponding to the frequency i.
Read ₁ p a (i) from the temporary coordinate storage unit 224, the temporary coordinate q ₁ p (i) the minimum coordinate value holding register 54
Is smaller than the value q _min retained in, q ₁
The value of p (i) is set to q _min . Also, the temporary coordinates q ₁
When the value of p (i) is larger than the value q _max held in the maximum coordinate value holding register 55, the value of q ₁ p (i) is set to q _max .

In the coordinate quantization step conversion processing S67, the same processing as the coordinate quantization step conversion processing S13 is performed. That is, referring to FIG. 2, the temporary coordinate q ₁ p (i) corresponding to the frequency i stored in the temporary coordinate storage device 224 is converted into a quantization step corresponding to the frequency i. Coordinate quantization step converter 24
Stores the value of the temporary coordinate q ₁ p (i) in the temporary coordinate storage device 22.
4 and the temporary coordinates q ₁ p (i) and 255.
A product with 0 is calculated, and the result is output to the quantization table temporary storage 25 and stored.

In the quantization table reading process S68, the same process as the above-described quantization table reading process S14 is performed. That is, the quantization table reading device 26 reads the value of the quantization step after the conversion in the coordinate quantization step conversion process S67 from the quantization table temporary storage device 25, and stores the value of the quantization step in the quantization table storage unit 2. Output to Next, in image evaluation processing S69, processing similar to the above-described image evaluation processing S15 is performed. That is, JPEG encoding processing S31,
A JPEG decoding process S32, a compression ratio calculation process S33 for calculating the compression ratio C _1n, and an image quality evaluation process S34 for calculating the image quality evaluation value C _2n are sequentially executed.

The evaluation function calculation processing S70 is the same processing as the evaluation function calculation processing S16 described above. In the evaluation function calculation processing S70, referring to FIG. 3, the updated evaluation function value U _new is calculated using the compression ratio C _1n and the image quality evaluation value C _2n calculated in the image evaluation processing S69. Specifically, the evaluation function calculating device 218 reads out the compression ratio C _1n from the compression ratio storage device 217 and the image quality evaluation value C _2n from the image quality evaluation value storage device 221, and based on the two values, Is used to calculate the updated evaluation function value U _new , and the result value is output to the updated evaluation function storage device 222. The image quality evaluation value C _2n is a color difference between the object color of the compressed image and the object color of the decoded image, but may be the SNR between the two images.

In the force calculation processing S71, referring to FIG. 3, the force calculation device 220 updates the pre-update evaluation function value U _old from the pre-update evaluation function value storage device 219 and updates the evaluation function value from the post-update evaluation function value storage device 222. The post-evaluation function value U _new and the minute coordinate difference dq (i) corresponding to the frequency i are read from the minute coordinate difference storage device 216, the force f is calculated using the following equation (19), and the resulting value is expressed as the force The data is output to the storage device 223 and stored.

[0087]

(19) f ← − (U _new −U _old ) / dq (i)

In the coordinate difference calculation processing S 72,
, The coordinate difference calculating device 29 calculates the coordinates q ₁ (i) corresponding to the frequency i from the coordinate storage device 23 and the pre-update coordinates q corresponding to the frequency i from the pre-update coordinate storage device 28.
₀ (i), the minute time constant dt from the minute time constant storage device 211, and the force f from the force storage device 223, and the coordinate difference diffQ corresponding to the frequency i is calculated using the following equation (20).
2Q1 (i) is calculated, and the result is output to the coordinate difference storage device 210.

[0089]

[Equation 20] diffQ2Q1 (i) ← ｛q ₁ (i) −q
₀ (i)｝ + dt × dt × f

In the speed update processing S73, referring to FIG.
And the speed v (i) corresponding to the frequency i and the force storage device 2
23, the force f, the minute time constant storage device 211, the minute time constant dt, and the attenuation constant storage device 215, the attenuation constant.
The dumping is read out, the speed is updated using the following equation (21), and the result is output to the speed storage device 213 and stored.

[0091]

## EQU21 ## v (i) ← v (i) + f × dt × dumping

In the process S74 for incrementing the frequency counter 58, the value of the frequency i held in the frequency counter 58 shown in FIG. 6 is incremented by one.
After setting the result value in the frequency counter 58, the process returns to step S62.

Next, as shown in FIG. 14, the updating process S18 includes a third repetitive control initialization process S81, a step S82 for determining the end, and a coordinate value updating process S83 before updating.
, A coordinate value update process S84, a reflection process S85, a coordinate range optimization process S86, and an increment process S87 of the frequency counter 58.

First, in the third repetitive control initialization processing S81, referring to FIG. 6, the control circuit 52 sets the frequency i of the frequency counter 58 to 1. Next, in step S82, the arithmetic circuit 51 sets the frequency counter 52
_Is compared with the maximum frequency i _max of the maximum frequency holding register 59, and if the ending condition of the following expression 22 is not satisfied, the process proceeds to step S83. On the other hand, if the end condition is satisfied, the control circuit 52 ends the update processing S18 and returns to the original routine.

[0095]

I> i _max Therefore, each processing S83 to S87 in the repetitive processing is performed.
Is performed on coordinate values and the like corresponding to all frequencies.

In the pre-update coordinate value updating process S83,
The control device 21 uses the coordinates q ₁ (i) stored in the coordinate storage device 23 and calculates the pre-update coordinates q ₀ (i) corresponding to each frequency i stored in the pre-update coordinate storage device 28 according to the following equation. ) To update.

## EQU23 ## q ₀ (i) ← q ₁ (i)

Next, in the coordinate value update processing S84, referring to FIG. 2, the coordinate calculation device 27 reads the coordinates q ₁ (i) corresponding to each frequency i from the coordinate storage device 23,
Reads the coordinate difference diffQ2Q1 (i) corresponding to the frequency i from the coordinate difference storing unit 210, and updates the coordinates q ₁ corresponding to each frequency i using the following equation (i), its result value coordinate storage device 23 And store it.

## EQU24 ## q ₁ (i) ← q ₁ (i) + diff Q2Q1 (i)

Next, in the reflection processing S85, if necessary depending on the value of the updated coordinates q ₁ (i) corresponding to the frequency i, the updated coordinates q ₁ (i) are:
The values of the speed v (i) and the pre-update coordinates q ₀ (i) are updated. The details of the reflection processing S85 will be described later. further,
In the coordinate range optimizing process S86, referring to FIG. 2, similarly to the coordinate range optimizing process S66, the controller 21
Reads the coordinate q ₁ (i) corresponding to the frequency i from the coordinate storage device 23, and stores the value in the minimum coordinate holding register 54.
_Is compared with the value q _min held at
₁ (i) is smaller than the value q _{mi n} is, q
₁ Set the value of (i) to q _min . Also, the coordinates q ₁ (i)
The value compared with the value q _max held in the maximum coordinate value holding register 55 of, if the coordinate q ₁ (i) is greater than the value q _max is, q ₁ and the value of (i) to q _max Set.

In step S87 for incrementing the frequency counter 58, the value of the frequency i held in the frequency counter 58 of FIG. 6 is incremented by one.
After setting the result value in the frequency counter 58, the process returns to step S82.

Here, the reflection processing S85 will be described in detail. As shown in FIGS. 15 and 16, the reflection process S85 includes a first comparison process S91, a first coordinate value justification process S92, a first speed reflection process S93, and a second coordinate value justification process. Processing S94, second comparison processing S95, and third processing
Coordinate value justification processing S96 and second velocity reflection processing S9
7 and a fourth coordinate value justification process S98.

First, in the first comparison processing S91,
The arithmetic circuit 51 in the control unit 21 reads from the coordinate q ₁ (i) the coordinate storage unit 23 corresponding to the frequency i, the maximum coordinate values the coordinates q ₁ the value of (i) is held in the maximum coordinate value holding register 55 Compare with q _max . If the following equation is satisfied, the control circuit 52 shifts the next process to the first coordinate justification process S92. On the other hand, if the following expression is not satisfied, the control circuit 52 executes the next process in the second comparison process S95.
Transitions to.

[0102]

## EQU25 ## q ₁ (i)> q _max

In the first coordinate value justification process S92, referring to FIG. 6, the arithmetic circuit 51 in the control device 21 reads the value of the coordinate q ₁ (i) corresponding to the frequency i from the coordinate storage device 23, The maximum coordinate value q _max is read from the maximum coordinate value holding register 55, the value of the coordinate q ₁ (i) corresponding to the frequency i is updated using the following equation, and the coordinate storage device 2
3 and store it.

[0104]

[Equation 26] q ₁ (i) ← q _max − {q ₁ (i) −q _max }

Next, in the first speed reflection processing S93, the control device 21 reads the speed v (i) corresponding to the frequency i from the speed storage device 213 and the constant value absorption from the speed absorption constant holding register 510. The arithmetic circuit 52 in the control device 21 calculates the value of the speed v (i) corresponding to the frequency i using the following equation, and the control circuit 51 outputs the result value to the speed storage device 213 and stores it.

[0106]

(27) v (i) ← −v (i) × absorption

Further, in the second coordinate value justification processing S94, the control device 21 sends the coordinates q ₁ (i) corresponding to the frequency i from the coordinate storage device 23 and the speed storage device 213
, The speed v (i) corresponding to the frequency i and the minute time constant dt from the minute time constant storage device 211 are read, and the pre-update coordinates q ₀ (i) corresponding to the frequency i are calculated using the following equation: The result value is output to the pre-update coordinate storage device 28 and stored.

[0108]

(28) q ₀ (i) ← q ₁ (i) −v (i) × dt

[0109] Then, in the second comparison processing S95 of FIG. 16, the arithmetic circuit 52 in the control unit 21 reads from the coordinate q ₁ (i) the coordinate storage unit 21 corresponding to the frequency i, the coordinates q ₁ ( The value of i) is compared with the minimum coordinate value q _min held in the minimum coordinate value holding register 54.
If the following equation is satisfied, the control circuit 52 shifts the next processing to the third coordinate justification processing S96. On the other hand, if the following expression is not satisfied, the control circuit 52 ends the reflection processing S85 and returns to the original routine.

[0110]

## EQU29 ## q ₁ (i) <q _min

In the third coordinate value justification process S96, the arithmetic circuit 51 of the control device 21 determines the minimum coordinate value q _min from the minimum coordinate value holding register 54 and the coordinate value corresponding to the frequency i from the coordinate storage device 23. q ₁ (i) is read out, the value of the coordinate q ₁ (i) at the frequency i is updated using the following equation, and the control device 21 outputs the result value to the coordinate storage device 23 and stores it.

[0112]

[Equation 30] q ₁ (i) ← q _min + {q _min −q ₁ (i)}

Next, in the second velocity reflection processing S97, similarly to the first velocity reflection processing S93, the controller 2
1 reads the speed v (i) corresponding to the frequency i from the speed storage device 213 and the constant value absorption from the speed absorption constant holding register 510, and calculates the speed v corresponding to the frequency i using the following equation. The value of (i) is updated, and the result value is output to the speed storage device 213 and stored.

[0114]

[Equation 31] v (i) ← −v (i) × absorption

Further, in the fourth coordinate value justification process S98, similarly to the second coordinate value justification process S94, the control device 21 sends the coordinates q ₁ (i) corresponding to the frequency i from the coordinate storage device 23. ) And the frequency i from the speed storage device 213.
And the minute time constant storage device 21 corresponding to the speed v (i) corresponding to
The respective values of the minute time constant dt are read from 1 and the pre-update coordinates q ₀ (i) corresponding to the frequency i are updated using the following equation, and the result values are output to the pre-update coordinate storage device 28 and stored. I do. Then, the process returns to the original routine.

[0116]

(32) q ₀ (i) ← q ₁ (i) −v (i) × dt

In the above embodiment, a case has been described in which a quantization table having both a luminance quantization table and a chrominance quantization table is handled as a series of tables. However, the present invention is not limited to this. The table may have only a quantization table for luminance or only a quantization table for color difference.

In the above embodiment, a case has been described in which the quantization table is normalized by normalizing the coordinates to non-negative real numbers, and optimization is performed. However, natural numbers are handled as in a standard JPEG quantization table. May be configured to execute the optimization of the quantization table. Also,
As a modified example of the present embodiment, an MPEG encoding unit may be used instead of the JPEG encoding unit 3, and an MPEG encoding unit may be used instead of the JPEG decoding unit 4.

The processing program for optimizing the quantization table described in the above embodiment may be implemented by, for example, a CD-RO
The information may be stored on a recording medium such as an optical disk such as M, CD-R, CD-RW, and DVD, or a magneto-optical disk, or a floppy disk. This makes it easy to provide a processing program for the quantization table optimizing process, and can be widely executed by a general-purpose computer. Further, the compression ratio may be a bit rate (amount of information expressed in bits) which is the number of bits per pixel of the decoded image data.

[0120]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A part of the result of the quantization table optimizing process performed as described above will be described. The test image to be compressed used here is an ITE standard pattern "hairband woman" which is a digital standard image of the Institute of Television Engineers of Japan. R, G, and B images in the “Sunraster” format (an image format proposed by Sun Microsystems, USA). This is an image cut out at a position where the x coordinate is 0 pixel and the y coordinate is 0 pixel as a reference and the width is 704 pixels and the height is 480 pixels.

As the image quality evaluation value, the color difference of the object color is used.
When measuring the color difference of the object color, the XY
In conversion to the Z color space, gamma correction is performed by raising the values of R, G, and B to the power of 2.2, and tristimulus values in a field of view twice using standard D65 light are used. The details are described in the prior art document “Applied Physics Society, Optical Society”, “Color Properties and Technology”, pp. 91-93 (1998).

When the quantization table for this test image is optimized using the above-described algorithm, the quantization table for the luminance signal is obtained as shown in FIG.
7 and the color difference signal quantization table is shown in FIG.
8, the compression ratio is higher than when the quantization table of the exemplary value of the JPEG encoding method is used, and the value of the color difference of the object color, which is the image quality evaluation value, is smaller. The absolute value of the evaluation function at increased. This is a case where the database search process S21 of the database device 225 is searched with the minimum image quality evaluation value, that is, the minimum color difference. In this case, the number of repetitions is 140. The compression rate when the optimized luminance signal quantization table and color difference signal quantization table are used is 6
It becomes 0.742 times, and the color difference of the object color becomes 5.526. On the other hand, when the quantization table of the example values of the JPEG encoding method is used, the compression ratio is 56.716 times, and the color difference of the object color is 5.670.

Here, FIG. 19 shows a quantization table for a luminance signal when each value of the quantization table is converted to an integer by rounding off the first decimal place, and similarly, a color difference signal when the value is converted to an integer. FIG. 20 shows a quantization table for use. When these quantization tables for the luminance signal and the color difference signal are used, the compression ratio is 60.710, and the color difference of the object color is 5.485. When compressing an image using these optimized quantization tables, a user can obtain a high compression rate and excellent image quality without expanding the existing JPEG decoder, that is, without changing it at all. Images can be used.

As described above, in the embodiment of the present invention,
By combining the above-described means and optimizing the quantization table for image compression by repeating the change of the value of the quantization table, it is applicable to more general image data. A quantization table that satisfies a higher compression rate and better image quality at the same time than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method is obtained. be able to.

[0125]

As described above in detail, the method for optimizing a quantization table in image encoding according to the first aspect of the present invention uses a quantization table to encode compressed image data to be encoded. An encoding step of encoding data; a decoding step of decoding the encoded image data into decoded image data using the quantization table; and a data of the compressed image data and the encoded image data. An image quality evaluation in which a compression ratio, which is a bit rate, which is a bit rate, which is the number of bits per pixel of decoded image data, is calculated, and the image quality of the decoded image data is evaluated in comparison with the compressed image data. Calculating a scale, calculating an evaluation function value based on the compression ratio and the image quality evaluation scale, and updating the quantization table based on the value of the evaluation function And updating step that, with the encoding step, and the decoding step, and the calculation step,
A control step of optimizing the quantization table by repeating each step of the updating step a predetermined number of times. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

According to the method for optimizing a quantization table for image coding according to the second aspect, in the method for optimizing a quantization table for image coding according to the first aspect, the image quality evaluation is performed. The scale is defined by Japanese Industrial Standards Z8730 and is a color difference between the object colors of the image of the compressed image data and the object colors of the image of the decoded image data. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

According to a third aspect of the present invention, there is provided a quantization table optimizing apparatus for image encoding according to the first aspect, wherein the image quality evaluation is performed. The scale is a signal-to-noise ratio when the maximum value of the compressed image data is defined as a signal and the difference between the compressed image data and the decoded image data is defined as noise. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

According to a fourth aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding according to any one of the first to third aspects. The evaluation function is defined by adding a value obtained by multiplying the compression ratio by a predetermined first weighting factor and a value obtained by multiplying the image quality evaluation scale by a predetermined second weighting factor. The step changes the value of the quantization table so as to increase the absolute value of the evaluation function. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained. Further, the processing of the adaptation method can be further simplified.

Further, according to the method for optimizing the quantization table for image encoding according to claim 7, the quantization table for image encoding according to one of claims 1 to 6 is used. In the optimization method, the updating step includes a step of calculating a value of an evaluation function calculated using the current quantization table used for the encoding and decoding, and a predetermined minute value to the value of the current quantization table. Is updated based on the amount of change between the evaluation function and the value of the evaluation function calculated by using the quantization table when is added. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained. Further, the processing of the adaptation method can be further simplified.

According to the quantization table optimizing apparatus for image coding according to claim 8 of the present invention, the compressed image data to be coded is coded into the coded image data using the quantization table. Encoding means, decoding means for decoding the encoded image data into decoded image data using the quantization table, the ratio of the data amount of the compressed image data and the encoded image data, Alternatively, a compression ratio which is a bit rate which is the number of bits per pixel of the decoded image data is calculated, and an image quality evaluation scale for evaluating the image quality of the decoded image data by comparing with the compressed image data is calculated. Calculating means, calculating means for calculating a value of an evaluation function based on the compression ratio and the image quality evaluation scale, and updating the quantization table based on the value of the evaluation function; Comprising the stage, and the decoding means, and said calculating means, and control means for optimizing the quantization table by repeating a predetermined number of times the processing of the updating means. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

According to the quantization table optimizing apparatus for image encoding according to the ninth aspect, in the quantization table optimizing apparatus for an image encoding according to the eighth aspect, the image quality evaluation is performed. The scale is defined by Japanese Industrial Standards Z8730 and is a color difference of an object color indicating a distance between the object color of the image of the compressed image data and the object color of the image of the decoded image data. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

Further, according to the apparatus for optimizing a quantization table for image coding according to claim 10, the apparatus for optimizing a quantization table for image coding according to claim 8 is characterized in that: The scale is a signal-to-noise ratio when the maximum value of the compressed image data is defined as a signal and the difference between the compressed image data and the decoded image data is defined as noise. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

According to the apparatus for optimizing a quantization table for image coding according to the eleventh aspect of the present invention,
Or the quantization table optimizing apparatus for image encoding according to 10, wherein the evaluation function is obtained by multiplying the compression ratio by a predetermined first weighting factor and the image quality evaluation scale by a predetermined second weighting factor. The updating means changes the value of the quantization table so as to increase the value of the evaluation function.
Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained. Further, the processing of the adaptation method can be further simplified.

Further, according to the apparatus for optimizing a quantization table for image coding according to claim 14, the quantization table for image coding according to one of claims 8 to 13 is provided. In the optimizing device, the updating unit may determine a value of the evaluation function calculated by the calculating unit using the current quantization table used for the encoding and decoding, and a value of the current quantization table. The current value of the quantization table is updated based on the amount of change between the evaluation function and the value of the evaluation function calculated by the calculation means using the quantization table obtained by adding the small value of. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained. Further, the processing of the adaptation method can be further simplified.

According to the recording medium of claim 15 of the present invention, each step included in the method of optimizing a quantization table for image encoding according to one of claims 1 to 7 is described. Record the processing program. Therefore, the present invention can be applied to more general image data.
It is possible to obtain a quantization table that simultaneously satisfies a higher compression ratio and better image quality as compared with a case where image data is encoded using a quantization table of an example value of the EG encoding method. Further, the processing of the adaptation method can be further simplified. Further, it is possible to easily provide a processing program for the optimization processing of the quantization table, and it can be widely executed by a general-purpose computer.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration example of an image evaluation processing system according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a first part of the high-dimensional algorithm operation unit 1 of FIG.

FIG. 3 is a block diagram showing a second part of the high-dimensional algorithm operation unit 1 of FIG.

FIG. 4 is a block diagram showing a configuration of a JPEG encoding unit 3 of FIG.

FIG. 5 is a block diagram illustrating a configuration of a JPEG decoding unit 4 of FIG. 1;

FIG. 6 is a block diagram showing a configuration of a control device 21 of FIG.

FIG. 7 is a flowchart illustrating a quantization table optimization process which is a main routine executed by the image processing system of FIG. 1;

8 is an initialization processing S11 which is a subroutine of FIG. 7;
It is a flowchart which shows.

9 is an image evaluation process S1 which is a subroutine of FIG.
6 is a flowchart illustrating a fifth example.

10 is a flowchart showing a JPEG encoding process S31 which is a subroutine of FIG.

FIG. 11 is a flowchart showing a JPEG decoding process S32 which is a subroutine of FIG. 9;

FIG. 12 is a partial update process S which is a subroutine of FIG. 7;
17 is a flowchart showing a first part of FIG.

FIG. 13 is a partial update process S which is a subroutine of FIG. 7;
17 is a flowchart showing a second part of No. 17.

FIG. 14 is an update process S18 as a subroutine of FIG. 7;
It is a flowchart which shows.

FIG. 15 is a reflection process S8 which is a subroutine of FIG.
5 is a flowchart showing a first part of FIG.

FIG. 16 is a reflection process S8 as a subroutine of FIG.
6 is a flowchart showing a second part of FIG.

17 is a diagram illustrating an example of a luminance signal quantization table optimized in the image processing system of FIG. 1;

18 is a diagram illustrating an example of a color difference signal quantization table optimized in the image processing system of FIG. 1;

FIG. 19 is a diagram illustrating an example of a quantization table for luminance signals in FIG. 17 that has been converted to an integer.

FIG. 20 is a diagram showing an example of a quantization table for color difference signals in FIG. 18 converted into an integer;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... High-dimensional algorithm operation part, 2 ... Quantization table storage part, 3 ... JPEG encoding part, 4 ... JPEG decoding part, 5 ... Compressed image storage part, 6 ... JPEG image storage part, 7 ... Decoded image Storage unit, 8: image compression ratio calculation unit, 9: image quality evaluation unit, 21: control unit, 22: coordinate initial value storage unit, 23: coordinate storage unit, 24: coordinate quantization step conversion unit, 25: quantization Table temporary storage device, 26: quantization table reading device, 27: coordinate calculation device, 28: coordinate storage device before update, 29: coordinate difference calculation device, 210: coordinate difference storage device, 211: minute time constant storage device, 212 … Speed calculation device, 213 speed storage device, 214 minute coordinate difference calculation device, 215 attenuation coefficient storage device, 216 minute coordinate difference storage device, 217 compression ratio storage device 218 ... Evaluation function operation device, 219 ... Evaluation function value storage device before update, 220 ... Force operation device, 221 ... Image quality evaluation value storage device, 222 ... Evaluation function value storage device after update, 223 ... Power storage device 224: temporary coordinate storage device, 225: database device 31: compressed image reading device, 32: color conversion device, 33: pixel thinning device, 34: DCT operation device, 35: quantization device, 36: Huffman coding device, 37 JPEG image reading device 41 JPEG image reading device 42 Huffman decoding device 43 Dequantization device 44 IDCT operation device 45 Pixel interpolation device 46 Color conversion device 47 Decoding Image reading device, 51: arithmetic circuit, 52: control circuit, 53: storage circuit, 54: minimum coordinate value holding register, 55: maximum position Value holding register, 56 ... time counter, 57 ... maximum time holding register, 58 ... frequency counter, 59 ... maximum frequency holding register, 510 ... speed absorption constant hold register, 511 ... internal bus.

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] January 27, 1999

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] Claims

[Correction method] Change

[Correction contents]

[Claims]

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0008

[Correction method] Change

[Correction contents]

[0008]

According to a first aspect of the present invention, there is provided a method for optimizing a quantization table for image encoding, comprising the steps of: setting a first quantization table using a predetermined first initial value; A first initialization step of initializing, a second initialization step of initializing a second quantization table for encoding and decoding using a predetermined second initial value, and an evaluation function operation Repeating a step, a first update step, a second update step, the evaluation function calculation step, the first update step, and the second update step a predetermined number of times. And a control step of optimizing a second quantization table by using the second quantization table. The evaluation function operation step encodes the compressed image data to be encoded into encoded image data using a second quantization table. Coded image The data is decoded into decoded image data using the second quantization table, and the ratio of the amount of data between the compressed image data and the encoded image data, or the number of bits per pixel of the decoded image data Calculates a compression rate that is a bit rate that is, and calculates an image quality evaluation scale that evaluates the image quality of the decoded image data in comparison with the compressed image data, and calculates the calculated compression ratio and image quality evaluation scale. , The value (U _old ) of the first evaluation function is calculated, and the first updating step includes the step of updating the value (q1 (i)) of a certain element of the second quantization table to a predetermined minute value (q1 (i)). d
q (i)) to calculate a third quantization table, and encode the compressed image data to be encoded into encoded image data using the third quantization table. Is decoded into decoded image data using the third quantization table, and is the ratio of the data amount of the compressed image data to the encoded image data or the number of bits per pixel of the decoded image data. Calculates a compression rate that is a bit rate, and calculates an image quality evaluation scale that evaluates the image quality of the decoded image data by comparing with the compressed image data, based on the calculated compression rate and image quality evaluation scale. second evaluation function value (U _{ne w)} calculated Te,
A change amount between the calculated value of the first evaluation function (U _old ) and the calculated value of the second evaluation function (U _new ) is calculated, and the calculated change amount and the second change amount are calculated. On the basis of the value (q0 (i)) of the element of the first quantization table corresponding to the value (q1 (i)) of the element of the quantization table, the value (q0 (i)) of the element of the second quantization table q1 (i)) and the value of a predetermined variable (v (i)) corresponding to the value of the element are updated independently of the value of the element other than the element. The updating step is repeatedly executed for all the elements of the second quantization table, and the first
The second step having all elements for which the update step of
Is updated as the first quantization table.

[Procedure amendment 3]

[Document name to be amended] Statement

[Correction target item name] 0011

[Correction method] Change

[Correction contents]

According to a fourth aspect of the present invention, there is provided a quantization table optimizing apparatus for image encoding according to the first, second, or third aspect. The evaluation function is a value obtained by multiplying the compression ratio by a predetermined first weighting factor and the image quality evaluation scale by a predetermined second weighting coefficient.
And a value multiplied by a weighting factor of the first function is added, and the first updating step changes a value of a second quantization table so as to increase an absolute value of the evaluation function. I do. Here, when the evaluation function is constituted by the compression ratio and the color difference between the object colors, the first weighting coefficient and the second weighting coefficient are different codes. Further, when the evaluation function is composed of a compression ratio and a signal-to-noise ratio, the first weighting coefficient and the second weighting coefficient have the same sign.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0012

[Correction method] Deleted

[Procedure amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0013

[Correction method] Change

[Correction contents]

According to a seventh aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding, comprising: a first quantization table for initializing a first quantization table using a predetermined first initial value; Initialization means for initializing a second quantization table for encoding and decoding using a predetermined second initial value;
Each of the initialization means, the evaluation function calculation means, the first update means, the second update means, the evaluation function calculation means, the first update means, and the second update means Control means for optimizing the second quantization table by repeating the processing a predetermined number of times, wherein the evaluation function calculating means uses the second quantization table to encode the compressed image data to be encoded. Encoding into encoded image data, decoding the encoded image data into decoded image data using a second quantization table, and a ratio of a data amount between the compressed image data and the encoded image data, or Calculate a compression rate that is a bit rate that is the number of bits per pixel of the decoded image data, and calculate an image quality evaluation scale that evaluates the image quality of the decoded image data in comparison with the compressed image data; the above First evaluation function value (U _old) calculated based on the calculated compression ratio and image quality evaluation measure, the first update means, the value of certain elements of the second quantization table (q1 ( A third quantization table is calculated by adding a predetermined minute value (dq (i)) to (i)), and the compressed image data to be encoded is encoded using the third quantization table. The encoded image data is decoded into decoded image data using a third quantization table, and the ratio of the data amount between the compressed image data and the encoded image data or the decoded image data is encoded. Calculating a compression ratio, which is a bit rate that is the number of bits per pixel of the data, and calculating an image quality evaluation scale for evaluating the image quality of the decoded image data by comparing with the compressed image data;
Based on the calculated compression ratio and image quality evaluation scale, a second
Calculates the value of the evaluation function (U _{ne w),} the amount of change between the computed second evaluation function value which is calculated from the first evaluation function value (U _old) (U _new) Calculated, and the calculated change amount and the value (q1) of a certain element in the second quantization table.
Based on the value (q0 (i)) of the element in the first quantization table corresponding to (i)), the value (q1 (i)) of the element in the second quantization table and the value of the element in the second quantization table And updates the value of the predetermined variable (v (i)) corresponding to the second variable independently of the value of the element other than the element, and the second updating means performs processing of the first updating means by a second quantization. The process is repeatedly executed for all elements of the table, and the second quantization table having all the elements on which the processing of the first updating unit has been executed is updated as the first quantization table.

[Procedure amendment 6]

[Document name to be amended] Statement

[Correction target item name] 0014

[Correction method] Change

[Correction contents]

According to an eighth aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding according to the seventh aspect of the present invention. , Defined by Japanese Industrial Standards Z8730, and is a color difference of an object color indicating a distance between the object color of the image of the compressed image data and the object color of the image of the decoded image data.

[Procedure amendment 7]

[Document name to be amended] Statement

[Correction target item name] 0015

[Correction method] Change

[Correction contents]

Further, the apparatus for optimizing a quantization table for image coding according to claim 9 is the apparatus for optimizing a quantization table for image coding according to claim 7.
The image quality evaluation scale is a signal-to-noise ratio when the maximum value of the compressed image data is a signal, and the difference between the compressed image data and the decoded image data is noise.

[Procedure amendment 8]

[Document name to be amended] Statement

[Correction target item name] 0016

[Correction method] Change

[Correction contents]

According to a tenth aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding according to the seventh, eighth or ninth aspect.
In the above-described apparatus for optimizing a quantization table for image encoding, the evaluation function may include a value obtained by multiplying the compression ratio by a predetermined first weighting factor, and a value obtained by multiplying the image quality evaluation scale by a predetermined second weight. It is defined by adding a value multiplied by a coefficient, and the first updating means changes the value of the second quantization table so as to increase the absolute value of the evaluation function. Here, when the evaluation function is constituted by the compression ratio and the color difference between the object colors, the first weighting coefficient and the second weighting coefficient are different codes. When the evaluation function is constituted by a compression ratio and a signal-to-noise ratio, the first weighting factor and the second
And the weighting coefficient is set to the same sign.

[Procedure amendment 9]

[Document name to be amended] Statement

[Correction target item name] 0017

[Correction method] Deleted

[Procedure amendment 10]

[Document name to be amended] Statement

[Correction target item name] 0018

[Correction method] Change

[Correction contents]

According to a thirteenth aspect of the present invention, there is provided a recording medium according to any one of the first to sixth aspects, wherein each of the steps is included in the method for optimizing a quantization table for image encoding. Is recorded.

[Procedure amendment 11]

[Document name to be amended] Statement

[Correction target item name] 0045

[Correction method] Change

[Correction contents]

The initialization process S11 is an initialization process for optimizing the quantization table stored in the quantization table storage unit 2, and performs a process such as setting the initial value of the time t to 1. . Optimization of the quantization table includes coordinate quantization step conversion processing S13, quantization table reading processing S14, image evaluation processing S15, evaluation function calculation processing S16, partial update processing S17, and update processing S18.
And the processing S19 of incrementing the time counter 56
And the respective steps of the database registration processing S20
It is performed by repeating until the termination condition of 2 is satisfied.

[Procedure amendment 12]

[Document name to be amended] Statement

[Correction target item name] 0047

[Correction method] Change

[Correction contents]

[0047]

T> t _max Therefore, each of the processes S13 to S20 in the iterative process is t
Executed _max times.

[Procedure amendment 13]

[Document name to be amended] Statement

[Correction target item name] 0054

[Correction method] Change

[Correction contents]

Next, the partial update process S17 is a process of calculating a coordinate difference or the like for calculating a coordinate value to be updated for each frequency by a high-dimensional algorithm. The update process S18 is a process of updating the coordinate value for each frequency using the coordinate difference calculated in the partial update process S17. In the database registration process S20, the control device 21 transmits the quantization table at time t,
The compression ratio C ₁ , the image quality evaluation value C _2, and the evaluation function value are calculated at time t.
At the same time, a process of registering in the database device 225 is performed. Further, in the process S19 of incrementing the time counter 56, the value of the time t stored in the time counter 56 is incremented by 1, and after performing the database registration process S20, the process returns to step S12. When the end condition of step S12 is satisfied, a database search process S21 is performed. Database search processing S21
The control unit 21 uses the maximum compression rate, the minimum image quality evaluation value when the image quality evaluation value is a color difference, the maximum image quality evaluation value or the maximum evaluation function value when the image quality evaluation value is a signal-to-noise ratio as a search key. A process of searching the database device 225 for a desired quantization table is performed. Here, it is assumed that a desired maximum compression ratio or the like serving as a search key is set in advance by the user.

[Procedure amendment 14]

[Document name to be amended] Statement

[Correction target item name] 0125

[Correction method] Change

[Correction contents]

[0125]

As described above in detail, according to the method for optimizing a quantization table in image encoding according to the first aspect of the present invention, the first quantization table is set to a predetermined first initial value. A first initialization step of initializing using a first quantization step, and a second quantization table for encoding and decoding.
A second initialization step of initializing using an initial value of
An evaluation function calculation step, a first update step, and a second
, The evaluation function calculation step, the first update step, and the second update step are repeated a predetermined number of times to optimize the second quantization table. Wherein the evaluation function calculating step encodes the compressed image data to be encoded into encoded image data using a second quantization table, and converts the encoded image data into a second quantization table. The decoded image data is used to decode the compressed image data, and the ratio of the data amount of the compressed image data to the encoded image data or the compression rate which is the bit rate which is the number of bits per pixel of the decoded image data is calculated. Calculating an image quality evaluation scale for evaluating the image quality of the decoded image data by comparing the image quality with the compressed image data, and calculating the calculated compression ratio and image quality evaluation scale First evaluation function values based on
(U _old ), and the first updating step is to add a predetermined minute value (dq (i)) to a value (q1 (i)) of a certain element in the second quantization table. A third quantization table is calculated, the compressed image data to be encoded is encoded into encoded image data using the third quantization table, and the encoded image data is encoded using the third quantization table. Decoding into decoded image data, calculating the ratio of the amount of data between the compressed image data and the encoded image data, or the compression rate which is the bit rate which is the number of bits per pixel of the decoded image data, And calculating an image quality evaluation scale that evaluates the image quality of the decoded image data in comparison with the compressed image data, and based on the calculated compression ratio and image quality evaluation scale, a value (U _{ne w} ) A change amount between the calculated first evaluation function value (U _old ) and the calculated second evaluation function value (U _new ) is calculated, and the calculated change amount and the second
Based on the value (q0 (i)) of the element of the first quantization table corresponding to the value (q1 (i)) of the element of the quantization table (Q
1 (i)) and a predetermined variable (v
The value of (i)) is updated independently of the values other than the element,
In the second updating step, the first updating step is repeatedly executed for all the elements of the second quantization table, and the second quantization step having all the elements for which the first updating step has been executed is performed. The quantization table is updated as the first quantization table. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

[Procedure amendment 15]

[Document name to be amended] Statement

[Correction target item name] 0128

[Correction method] Change

[Correction contents]

According to a fourth aspect of the present invention, there is provided an apparatus for optimizing a quantization table for image encoding according to any one of the first to third aspects. The evaluation function is defined by adding a value obtained by multiplying the compression ratio by a predetermined first weighting factor and a value obtained by multiplying the image quality evaluation scale by a predetermined second weighting factor. The first updating step changes the value of the second quantization table so as to increase the absolute value of the evaluation function. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.
Further, the processing of the adaptation method can be further simplified.

[Procedure amendment 16]

[Document name to be amended] Statement

[Correction target item name] 0129

[Correction method] Deleted

[Procedure amendment 17]

[Document name to be amended] Statement

[Correction target item name] 0130

[Correction method] Change

[Correction contents]

According to the apparatus for optimizing a quantization table for image encoding according to claim 7 of the present invention, the first quantization table is initialized using a predetermined first initial value. 1 initialization means, 2nd initialization means for initializing a second quantization table for encoding and decoding using a predetermined second initial value, evaluation function calculation means, The first updating means, the second updating means, the evaluation function calculating means, the first updating means, and the second updating means are each repeated a predetermined number of times to obtain a second updating means. Control means for optimizing the quantization table, wherein the evaluation function calculating means encodes the compressed image data to be encoded into encoded image data using the second quantization table,
The encoded image data is decoded into decoded image data by using a second quantization table, and the ratio of the data amount between the compressed image data and the encoded image data, or per pixel of the decoded image data, Calculates a compression rate, which is a bit rate that is the number of bits, and calculates an image quality evaluation scale for evaluating the image quality of the decoded image data in comparison with the compressed image data. A value (U _old ) of a first evaluation function is calculated based on the evaluation scale, and the first updating unit calculates a value (q1 (i)) of a certain element of the second quantization table as a predetermined minute value. Value (dq
(I)) is added to calculate a third quantization table, the compressed image data to be encoded is encoded into encoded image data using the third quantization table, and the encoded image data is A third quantization table is used to decode the decoded image data into decoded image data, and the ratio of the data amount of the compressed image data to the encoded image data or the number of bits per pixel of the decoded image data Calculates a compression rate that is a rate, and calculates an image quality evaluation scale that evaluates the image quality of the decoded image data in comparison with the compressed image data, based on the calculated compression rate and image quality evaluation scale. calculates a second evaluation function value (U _{ne w),} the change between the computed first evaluation function value (U _old) and a second evaluation function values that have been computed (U _{new new)} Calculate the quantity,
Based on the calculated change amount and the value (q0 (i)) of the element of the first quantization table corresponding to the value (q1 (i)) of a certain element of the second quantization table, In the quantization table of No. 2, the value of the element (q1 (i)) and the value of the predetermined variable (v (i)) corresponding to the value of the element are updated independently of the values other than the element. The second updating means is:
The processing of the first updating unit is repeatedly executed for all elements of the second quantization table, and the second quantization table having all the elements for which the processing of the first updating unit has been executed is stored in the second quantization table. It is updated as one quantization table.
Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

[Procedure amendment 18]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Change

[Correction contents]

According to the quantization table optimizing apparatus for image coding described in claim 8, in the quantization table optimizing apparatus for image coding described in claim 7, the image quality evaluation is performed. The scale is defined by Japanese Industrial Standards Z8730 and is a color difference of an object color indicating a distance between the object color of the image of the compressed image data and the object color of the image of the decoded image data. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

[Procedure amendment 19]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Change

[Correction contents]

Further, according to the quantization table optimizing apparatus for image encoding according to the ninth aspect, in the quantization table optimizing apparatus for image encoding according to the seventh aspect, the image quality evaluation is performed. The scale is a signal-to-noise ratio when the maximum value of the compressed image data is defined as a signal and the difference between the compressed image data and the decoded image data is defined as noise. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained.

[Procedure amendment 20]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Change

[Correction contents]

According to the quantization table optimizing apparatus for image coding according to the tenth aspect, the seventh and eighth aspects are described.
Or the quantization table optimizing device for image encoding according to 9, wherein the evaluation function is obtained by multiplying the compression ratio by a predetermined first weighting factor and the image quality evaluation scale by a predetermined second weighting factor. And the first updating means changes the value of the second quantization table so as to increase the value of the evaluation function. Therefore, the present invention can be applied to more general image data, and has a higher compression ratio and higher compression rate than when image data is encoded using a quantization table of an exemplary value of the conventional JPEG encoding method. A quantization table that simultaneously satisfies good image quality can be obtained. Also,
Processing of the adaptation method can be further simplified.

[Procedure amendment 21]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Deleted

[Procedure amendment 22]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Change

[Correction contents]

According to the recording medium of the thirteenth aspect of the present invention, each step included in the method of optimizing a quantization table for image encoding according to one of the first to sixth aspects is described. Record the processing program. Therefore, the present invention can be applied to more general image data.
It is possible to obtain a quantization table that simultaneously satisfies a higher compression ratio and better image quality as compared with a case where image data is encoded using a quantization table of an example value of the EG encoding method. Further, the processing of the adaptation method can be further simplified. Further, it is possible to easily provide a processing program for the optimization processing of the quantization table, and it can be widely executed by a general-purpose computer.

Continued on the front page (72) Inventor Junichi Yamada 5th Sanraya, Inaya, Koika-cho, Kyoto No. 5 Sanpani, Koya-cho, Kochi-cho, ATR Co., Ltd.

Claims (15)

[Claims] An encoding step of encoding compressed image data to be encoded into encoded image data using a quantization table; and converting the encoded image data into decoded image data using the quantization table. A decoding step of decoding, and calculating a compression ratio that is a bit rate that is a ratio of a data amount between the compressed image data and the encoded image data or a bit number per pixel of the decoded image data,
And calculating an image quality evaluation scale that evaluates the image quality of the decoded image data by comparing with the compressed image data; and calculating an evaluation function value based on the compression ratio and the image quality evaluation scale. Updating the quantization table based on the value of the evaluation function; repeating the encoding step, the decoding step, the calculation step, and the update step a predetermined number of times. And a control step of optimizing the quantization table according to (1). 2. The image quality evaluation scale is based on Japanese Industrial Standard Z8.
The image code according to claim 1, wherein the image code is a color difference of an object color defined by 730 and indicating a distance between the object color of the image of the compressed image data and the object color of the image of the decoded image data. Optimization method of quantization table for quantization. 3. The image quality evaluation scale is a signal-to-noise ratio when a maximum value of the compressed image data is a signal, and a difference between the compressed image data and the decoded image data is a noise. 2. The method of optimizing a quantization table for image encoding according to claim 1, wherein: 4. The evaluation function is defined by adding a value obtained by multiplying the compression ratio by a predetermined first weighting factor and a value obtained by multiplying the image quality evaluation scale by a predetermined second weighting factor. 4. The quantization table according to claim 1, wherein the updating step changes a value of the quantization table so as to increase an absolute value of the evaluation function. Optimization method. 5. The evaluation function according to claim 4, wherein the first weighting coefficient and the second weighting coefficient have different codes when constituted by a compression ratio and a color difference between object colors. A method for optimizing a quantization table for the described image coding. 6. The evaluation function according to claim 4, wherein when the compression function and the signal-to-noise ratio are used, the first weighting coefficient and the second weighting coefficient have the same sign.
A method for optimizing a quantization table for the described image coding. 7. The method according to claim 1, wherein the updating step includes: a value of an evaluation function calculated using the current quantization table used for the encoding and decoding; and a predetermined minute value added to the value of the current quantization table. 7. The value of the current quantization table is updated based on a change amount between the evaluation function and the value of the evaluation function calculated using the quantization table when. A method for optimizing a quantization table for image encoding according to one of the above. 8. An encoding means for encoding compressed image data to be encoded into encoded image data using a quantization table, and converting the encoded image data into decoded image data using the quantization table. Decoding means for decoding, and a compression ratio which is a bit rate which is a ratio of a data amount between the compressed image data and the encoded image data or a bit number per pixel of the decoded image data,
Calculating means for calculating an image quality evaluation scale that evaluates the image quality of the decoded image data by comparing with the compressed image data; calculating an evaluation function value based on the compression ratio and the image quality evaluation scale; Updating means for updating the quantization table based on the value of the evaluation function; repeating the processing of the encoding means, the decoding means, the calculating means, and the updating means a predetermined number of times. And control means for optimizing the quantization table according to the above. 9. The image quality evaluation scale is based on Japanese Industrial Standard Z8.
9. The image coding method according to claim 8, wherein the color difference is an object color defined by 730 and indicating a distance between the object color of the image of the compressed image data and the object color of the image of the decoded image data. Table Optimizer for Macintosh. 10. The image quality evaluation scale is a signal-to-noise ratio when a maximum value of the compressed image data is a signal and a difference between the compressed image data and the decoded image data is a noise. 9. The apparatus for optimizing a quantization table for image encoding according to claim 8, wherein 11. The evaluation function is defined by adding a value obtained by multiplying the compression ratio by a predetermined first weighting factor and a value obtained by multiplying the image quality evaluation scale by a predetermined second weighting factor. 11. The quantization table for image encoding according to claim 8, wherein the updating unit changes a value of the quantization table so as to increase an absolute value of the evaluation function. Optimization device. 12. When the evaluation function is constituted by a compression ratio and a color difference between object colors, a first weighting factor and a second weighting factor are used.
12. The quantization table optimizing apparatus for image encoding according to claim 11, wherein the weighting coefficient is different from the weighting coefficient. 13. When the evaluation function is constituted by a compression ratio and a signal-to-noise ratio, a first weighting factor and a second
12. The apparatus according to claim 11, wherein the weighting coefficient is the same as the weighting coefficient. 14. The updating means according to claim 1, wherein the value of the evaluation function calculated by the calculating means using the current quantization table used for the encoding and decoding, and the value of the current quantization table are predetermined. Updating the current quantization table value based on the amount of change between the evaluation function value calculated by the calculation means using the quantization table when adding a small value of An apparatus for optimizing a quantization table for image encoding according to any one of claims 8 to 13. 15. A recording medium recording a processing program of each step included in the method of optimizing a quantization table for image encoding according to claim 1. Description: