JP2003153275A - Image processing apparatus and method, recording medium, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate decoded image data with high image quality by decoding encoded original image data. SOLUTION: A decoded image generating adaptive filter 31 applies filter processing to original image data to produce predicted image data and outputs the data to a high-speed fractal encoding section 32. An adder 33 subtracts the predicted image data from the original image data, produces a residual component and outputs it to a residual component adaptive quantization section 34. The high-speed fractal encoding section 32 encodes prediction image data through fractal encoding processing and outputs compressed image data to a transmission format encoding section 35. A residual component adaptive quantization section 34 adaptively quantizes the residual component by a dynamic range of the predicted image data and outputs the result to the transmission format encoding section 35. The transmission format encoding section 35 converts the compressed image data and the quantized residual component into a transmission code and provides an output of the result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information processing apparatus and method, a recording medium, and a program, and more particularly to fractal coding processing and MPEG (Moving Picture Experts G).
The present invention relates to an information processing apparatus and method, a recording medium, and a program that enable fractal decoding of an image coded by a coding process with high image quality and suppress the amount of calculation required for the process.

[0002]

2. Description of the Related Art Techniques for compressing images are becoming popular. As a technique for compressing an image, fractal coding has been proposed in which image redundancy is performed by removing the redundancy of self-similarity (partial self-similarity) with a specific portion of an image. The partial self-similarity is a property that, when paying attention to an image pattern of a part of an image, similar image patterns exist in different scales in the same image. In fractal encoding, an image to be encoded is divided into m × n range blocks, and a conversion process is applied to each range block. In this conversion, the pixel values of appropriately determined domain blocks are extracted, and simple linear conversions such as reduction and rotation are applied to the pixel values, and the pixels in the range block are replaced. That is, fractal coding is a coding method that determines a transform that accurately approximates an image pattern of a range block.

More specifically, for example, one frame image is divided into 8 × 8 pixel range blocks, and 16
When a × 16 pixel domain block is set, four types of conversion elements (reduction conversion, rotation conversion, mirror image conversion, luminance conversion) are set, and the total including the conversion of these combinations is set. Eight types of transforms are used. The reduction conversion is to reduce the horizontal and vertical lengths to 1/2 by averaging the values of 4 pixels. Rotation conversion is 0
Counterclockwise rotation of 90 °, 180 °, 270 °. The mirror image conversion is a process of switching left and right. The brightness conversion is a process of multiplying the gain after the average value separation and reducing the gain in the amplitude direction.

After performing the above-mentioned transformation on the domain block, the similarity with the range block is found by block matching, and the position information of the domain block, that is, the vector and the parameter of the above equation (that is, the position within the search range, the rotation Angle, whether it was flipped left or right)
Is transmitted, the amount of information is compressed.

On the decoding side, a domain block is cut out from the initial image, and the result of conversion using the received parameters is pasted to the initial image as a range block,
The image is updated accordingly. By repeating this operation, a restored image with gradually increasing resolution is generated.

[0006]

However, since the compression process by fractal coding is a process that can be estimated by repeating geometric changes and block comparisons, the amount of calculation and the processing time are enormous. There was a problem that it would become a thing.

Further, in fractal coding, the quality of block matching greatly affects the image quality.
In the case of an image having a fine pattern or the like, there is a problem that the image quality of a decoded image is significantly deteriorated because matching between blocks to be compared fails.

Further, various measures have been taken at the time of encoding to solve this problem. Among them, for example, in Japanese Patent Laid-Open No. 9-139941, luminance normalization processing is performed by ADRC (Adaptive
It has been proposed to reduce the amount of calculation processing by reducing the 8-bit data to 4-bit data by Dynamic Range Coding), but the reduction of the image quality deterioration due to the fractal coding processing has not been realized.

The present invention has been made in view of the above circumstances, and is to enable an image coded by a compression process by fractal coding of image data to be decoded with higher image quality and at higher speed. .

[0010]

A first information processing apparatus according to the present invention comprises a generation means for generating, from input image data, predictive image data which is predicted to be decoded after being encoded. Encoding means for encoding image data, difference calculating means for calculating difference data between input image data and predicted image data, and difference data calculated by the difference calculating means correspond to the feature amount of the predicted image data. Output adaptively quantized quantizing means, predictive image data encoded by the encoding means, and differential data adaptively quantized with respect to the feature amount of the predictive image data as encoded data. And output means.

The encoding means may perform encoding by performing fractal encoding processing on the predicted image data.

The quantizing means is quantized to indicate to the output means that it is not necessary to output the residual component of the block as encoded data according to the value of the dynamic range for each block. Can be allowed to.

The quantizing means does not need to output the residual component of the block as encoded data when the value of the dynamic range is smaller than a predetermined threshold value according to the value of the dynamic range for each block. It may be quantized as shown to the output means.

The characteristic amount may be a dynamic range of pixel values for each block of the predicted image data.

For each block, the quantizing means compares the pixel value of the predicted image data with the value obtained by multiplying the dynamic range by a predetermined coefficient, and the difference calculating means calculates in accordance with the comparison result. The difference data can be adaptively quantized.

The predetermined coefficient may be adaptively calculated for each block of the predicted image data in accordance with the dynamic range.

A first information processing method according to the present invention comprises a step of generating predicted image data which is predicted to be decoded after being encoded from input image data, and to encode the predicted image data. The encoding step, the difference calculation step for calculating the difference data between the input image data and the predicted image data, and the difference data calculated in the processing of the difference calculation step are adaptively adjusted according to the feature amount of the predicted image data. Quantization step to quantize, prediction image data encoded by the processing of the encoding step, and output step of outputting difference data adaptively quantized to the feature amount of the prediction image data as encoded data It is characterized by including and.

The program of the first recording medium of the present invention is
A generation control step for controlling generation of predictive image data that is predicted to be decoded after being encoded from the input image data, an encoding control step for controlling encoding of the predictive image data, and the input image data And a difference calculation control step for controlling the calculation of difference data between the predicted image data and
Quantization control step that controls adaptive quantization of the difference data whose calculation is controlled by the processing of the difference calculation control step corresponding to the feature amount of the predicted image data, and coding is controlled by the processing of the coding control step. And output control step for controlling the output of the differential data adaptively quantized to the feature amount of the predicted image data as encoded data.

A first program of the present invention comprises a generation control step for controlling generation of predictive image data which is predicted to be decoded after being encoded from input image data, and encoding of the predictive image data. An encoding control step for controlling, a difference calculation control step for controlling calculation of difference data between the input image data and the predicted image data, and difference data whose calculation is controlled by the processing of the difference calculation control step,
For the quantization control step that controls adaptive quantization corresponding to the feature amount of the predictive image data, the predictive image data whose encoding is controlled by the process of the encoding control step, and the feature amount of the predictive image data An output control step of controlling the output of the adaptively quantized difference data as encoded data, and causing the computer to execute the output control step.

According to the second information processing apparatus of the present invention, when the original image data is encoded and then further decoded, the predicted image data predicted to be decoded is encoded image data. And the encoded data consisting of the difference data between the original image data and the predicted image data adaptively quantized with respect to the feature amount of the predicted image data, the encoded image data and the quantized difference data are separated. Separation means, decoding means for decoding the coded image data separated by the separation means to predictive image data, and quantized difference data are adaptively inversely quantized corresponding to the feature amount of the predicted image data. Dequantizing means for converting the predicted image data, generating means for generating decoded image data based on the dequantized difference data, and an output means for outputting the decoded image data generated by the generating means. Characterized in that it comprises and.

The decoding means may perform fractal decoding processing on the coded image data to decode the coded image data into predicted image data.

The feature amount may be a dynamic range of pixel values for each block of the predicted image data.

The dequantizing means adaptively dequantizes each block by multiplying the dynamic range by a predetermined coefficient and further corresponding to the value obtained by multiplying the quantized difference data. Can be

It is possible to further provide coefficient adaptive calculation means for adaptively calculating the predetermined coefficient for each block of the predicted image data in accordance with the dynamic range.

According to the second information processing method of the present invention, when the original image data is encoded and then further decoded, the predicted image data predicted to be decoded is encoded image data. And the encoded data consisting of the difference data between the original image data and the predicted image data adaptively quantized with respect to the feature amount of the predicted image data, the encoded image data and the quantized difference data are separated. A separation step,
The encoded image data separated in the separation step is
The decoding step of decoding into the prediction image data, and the quantized difference data, corresponding to the feature amount of the prediction image data,
Dequantization step of adaptively dequantizing, prediction image data, generation step of generating decoded image data based on the dequantized difference data, and decoded image data generated by the processing of the generation step And an output step for outputting.

The program of the second recording medium of the present invention is
When the original image data is encoded and then further decoded,
From the encoded image data obtained by encoding the predicted image data that is predicted to be decoded, and the difference data between the original image data and the predicted image data that are adaptively quantized with respect to the feature amount of the predicted image data. From the encoded data, the separation control step for controlling the separation of the encoded image data and the quantized difference data, and the encoded image data whose separation is controlled by the processing of the separation control step to the predicted image data. Decoding control step for controlling the decoding, quantized difference data, corresponding to the feature amount of the predicted image data, inverse quantization control step for controlling the adaptive inverse quantization, the predicted image data, A generation control step that controls generation of decoded image data based on the dequantized difference data, and an output that controls output of the decoded image generation of which is controlled by the processing of the generation control step. Characterized in that it comprises a control step.

The second program of the present invention is encoded image data in which predictive image data predicted to be decoded is encoded when the original image data is encoded and then further decoded, Controls separation of coded image data and quantized difference data from coded data consisting of difference data between original image data and predicted image data that are adaptively quantized to the feature amount of the predicted image data And a decoding control step for controlling decoding of encoded image data whose separation is controlled by the processing of the separation control step into predicted image data, and characteristics of predicted image data of quantized difference data Inverse quantization control step for controlling adaptive inverse quantization corresponding to the amount, and control of generation of decoded image data based on predicted image data and dequantized difference data. A generation control step, generating in the process of generation control step is characterized by executing an output control step of controlling the output of the controlled decoded image to the computer.

A first information processing apparatus and method of the present invention,
Also, in the program, predicted image data that is predicted to be decoded after being encoded is generated from the input image data, the predicted image data is encoded, and difference data between the input image data and the predicted image data is generated. Is calculated,
The calculated difference data is adaptively quantized corresponding to the feature amount of the predicted image data, and the encoded predictive image data and the difference adaptively quantized with respect to the feature amount of the predicted image data The data is output as encoded data.

A second information processing apparatus and method of the present invention,
In addition, in the program, when the original image data is encoded and then further decoded, the predicted image data predicted to be decoded is encoded image data and the feature amount of the predicted image data. On the other hand, the encoded image data and the quantized difference data are separated from the encoded data composed of the difference data between the adaptively quantized original image data and the predicted image data, and the separated encoded image data is , The predicted image data is decoded, and the quantized difference data is adaptively dequantized corresponding to the feature amount of the predicted image data, and based on the predicted image data and the dequantized difference data. The decoded image data is generated, and the generated decoded image data is output.

[0030]

1 is a diagram showing a configuration of an embodiment of an image transfer system 11 according to the present invention.

The image transfer system 11 decodes the transmission data of the original image data encoded by the encoding unit 21 and the encoding unit 21 which encodes the input original image data into transmission data to obtain a decoded image. It is composed of a decoding unit 22 that generates data.

The decoded image generation adaptive filter 31 of the image transfer system 11 is a filter generated by a learning process which will be described later, and stores the input original image data in ROM (Read On).
calculation is performed using a coefficient stored in a storage device such as a ly memory), converted into prediction image data, and output to the high-speed fractal coding unit 32, the adder 33, and the residual component adaptive quantization unit 34. . More specifically, in the decoded image generation adaptive filter 31, the input original image is coded by the high-speed fractal coding unit 32, and then further decoded by the fractal decoding unit 42 of the decoding unit 22 described later. At this time, the high-speed fractal coding unit 32, the adder 33,
And output to the residual component adaptive quantization unit 34.

The high-speed fractal encoder 32 encodes (compresses) the original image data calculated by the decoded image generation adaptive filter 31 by fractal encoding and outputs the encoded image data to the transmission format encoder 35. To do.

The adder 33 subtracts the pixel value of the predicted image data generated by being filtered by the decoded image generation adaptive filter 31 from each pixel value of the input original image data, and the difference is the residual difference. It is obtained as a component and is output to the residual component adaptive quantization unit 34.

The residual component adaptive quantizer 34 includes an adder 33.
The input residual component is decoded image generation adaptive filter 3
The predicted image data filtered by 1 is adaptively quantized for each predetermined block in accordance with the dynamic range that is the feature amount, and is output to the transmission format unit 35.

The transmission format section 35 includes the encoded image data compressed by the high-speed fractal encoding section 32,
The residual component quantized by the residual component adaptive quantizer 34 and the residual component are converted into a predetermined transmission format and output as transmission data.

The transmission parameter separation unit 41 of the decoding unit 22
Separates the input transmission data into coded image data and a quantized residual component, outputs the coded image data to the fractal coding unit 42, and converts the quantized residual component into a residual component. Output to the adaptive quantizer 43.

The fractal decoding unit 42 decodes the coded image data obtained by coding the predictive image data generated by the fractal coding process on the coding side, which is input from the transmission parameter separation unit 41, and the added data is sent to the adder 44. In addition to the output, the information of the dynamic range for each block of the decoded predicted image data is output to the residual component adaptive dequantization unit 43.

The residual component adaptive dequantization unit 43 adaptively dequantizes the quantized residual component input from the transmission parameter separation unit 41 with respect to the dynamic range input from the fractal decoding unit 42. To generate a residual component and output it to the adder 44. The adder 44 generates decoded image data corresponding to the original image data by adding the prediction image data input from the fractal decoding unit 42 and the residual component, and outputs the decoded image data to the clipping processing unit 45.

The clipping processing unit 45 confirms whether or not the pixel value of each pixel of the decoded image data is within the defined range, and when the pixel value is less than the lower limit value of the defined range, Corrects the pixel value to the lower limit value, and when the pixel value exceeds the upper limit value of the defined range, the pixel value is clipped by correcting the pixel value to the upper limit value, and the decoded image data is corrected and output. To do.

Next, the structure of the high-speed fractal coding unit 32 will be described with reference to FIG.

The original image data is converted into the range blocking circuit 6
1 and the search area cutout unit 65. The range blocking unit 61 divides the original image data into two-dimensional (m × n), for example, 8 × 8 pixel size range blocks, and the ADRC encoder 62 for each range block.
Supply to. The ADRC encoder 62 ADRC encodes the input range block by ADRC encoding processing,
Dynamic range DR for each block range and minimum value MIN
To the transmission format encoding unit 35 and the encoding result to the comparison unit 63. The ADRC encoding process detects the maximum pixel value MAX and minimum pixel value MIN for each block,
The dynamic range DR, which is the difference between X and the minimum value MIN, is obtained, and the modified input pixel value obtained by subtracting the minimum value MIN from the original quantization bit number ( Quantization is performed again with a fixed number of bits (for example, 4 bits) smaller than 8 bits. Dynamic range D, which is the result of encoding
R and the minimum value MIN are sent to the transmission format coding unit 35, and the quantized data Qr of each pixel is sent to the comparison unit 63 that calculates the sum of squared differences. As the information on the dynamic range, two of the dynamic range DR, the maximum value MAX, and the minimum value MIN may be transmitted.

On the other hand, the search area slicing unit 65 selects a two-dimensional (M ×) from the search range corresponding to the range block (for example, in order to reduce the calculation time, it is approximately −7 to +8 pixels in both horizontal and vertical directions). N), for example 16x16
A domain block having a size of about a pixel is cut out and output to the domain blocking unit 66. The domain block conversion unit 66 performs the same ADRC encoding processing as the ADRC encoder 62 on the extracted domain block by the ADRC encoder 67 and outputs the same to the 1/4 reduction unit 68, and at the same time, outputs the conversion parameter to the comparison unit 63 and the minimum value determination. Output to the unit 64.

The 1/4 reduction unit 68 is used for the quantized data in the encoded output of the domain block from the ADRC encoder 67.
The Qd is subjected to reduction conversion processing (1/2 in each of the vertical and horizontal directions) and output to the rotating mirror image processing unit 69. The rotating mirror image processing unit 69 rotates the reduced quantized data Qd (0
°, 90 °, 180 °, 270 °), mirror image (horizontal inversion)
Etc., and outputs to the comparison unit 63.

The processing performed in the 1/4 reduction section 68 and the rotating mirror image processing section 69 is coordinate conversion processing. The conversion parameters in the 1/4 reduction unit 68 and the rotating mirror image processing unit 69 are supplied to and stored in the comparison unit 63 and the minimum value determination unit 64.

The evaluation unit 63 calculates the difference (Qr-Qd ') in pixel units between the range block and the domain block, calculates the sum of squared values of this difference within the block, and calculates the calculated difference of 2 The sum of the power values is stored (stored) as an evaluation value table. Note that the sum of absolute differences or the like can be used as the evaluation value.

The evaluation unit 63 includes a 1/4 reduction unit 68, and
When the parameters are changed and the conversion processing is continuously executed in the rotating mirror image processing unit 69, the sum of the squared values of the above-mentioned difference is obtained and stored (stored) as an evaluation value table. The evaluation unit 63 performs this operation for all the parameters and finishes the calculation for the position of one domain block within the search range. As an example, the following eight types of parameters are used.

The first parameter is based on 1/2 reduction processing, the second parameter is based on 1/2 reduction and horizontal inversion processing, and the third parameter is 1/2.
Reduction and 90 ° rotation processing, the fourth parameter is 1/2 reduction and 180 ° rotation processing, and the fifth parameter is 1/2 reduction and 27
The 6th parameter is 1 /
2 reduction, 90 ° rotation and horizontal inversion processing, the 7th parameter is 1/2 reduction, 180 ° rotation and horizontal inversion, and the 8th parameter is 1 /
2 reduction, 270 ° rotation, and left / right inversion.

The search area cutout unit 65 to the rotating mirror image processing unit 69 cut out the domain block within the search range by shifting by one pixel, for example, and the evaluation unit 63.
Performs the same evaluation value calculation for all conversion parameters as described above, cuts out a domain block at a position moved by one pixel in the search range, and executes all calculations.

When the evaluation section 63 executes all the operations,
The minimum value determination unit 64 detects the minimum value from the evaluation values stored as an evaluation value table in the evaluation unit 63, and determines the position of the domain block and the conversion parameters (rotation angle, presence / absence of left / right inversion). , Transmission format encoder 3
Output to 5. In this way, an image similar to the range block is searched for by the block matching method. The transmission format encoding unit 35 uses the dynamic range DR of the range block supplied from the ADRC encoder 62,
Along with the minimum value MIN, the quantized data having the minimum evaluation value input from the minimum value determination unit 64 and the conversion parameter information are converted into a transmission format to generate transmission data.

The high-speed fractal coding section 32 executes the above series of coding operations for each range block, and performs the coding operation over the entire image frame. The transmitted data includes position information x and y of 4 bits, 3 bits of conversion parameters (2 bits for rotation and 1 bit for inversion), 8 bits of dynamic range DR,
The minimum value MIN is 8 bits. Therefore, when one pixel is 8-bit data, the original data of 8 × 8 × 8 = 512 bits is compressed to 27 bits.

The control unit 70 is composed of a so-called microcomputer or the like, controls the entire operation of the high-speed fractal encoding unit 32, and is provided with a counter required for various processes (connections are omitted in the figure). Exist).

Next, the configuration of the fractal decoding unit 42 will be described with reference to FIG.

The domain block cutout unit 81 cuts out a domain block (16 × 16 pixels) based on the cutout information (position information x, y) of the domain block supplied from the transmission parameter separation unit 41, and supplies it to the ADRC encoder 82. To do. The ADRC encoder 82 performs ADRC encoding processing on the data of the domain block and outputs the quantized data to the 1/4 reduction unit 83. The 1/4 reduction unit 83 horizontally quantizes the quantized data input from the ADRC encoder 82.
Then, the image is reduced by 1/2 in the vertical direction and output to the rotating mirror image processing unit 84.

The rotating mirror image processing unit 84 rotates, according to the conversion parameter supplied from the transmission parameter separating unit 41.
Linear conversion such as mirror image processing is performed on the quantized data of the domain block supplied from the 1/4 reduction unit 83, and ADRC
It is supplied to the decoder 85. The ADRC decoder 85 performs a decoding process on the converted quantized data from the rotating mirror image processing unit 84 based on the dynamic range DR and the minimum value MIN from the transmission parameter separating unit 41 to restore the original pixel information. , And writes it as the decoding result FM0 in the memory 86.

Similarly, the fractal decoding unit 42 executes the decoding operation of the encoded data corresponding to the next range block, and when the decoding is completed over the entire one frame, the iterative operation is performed again. To execute.
In the determination for converging this iterative operation, the previous decoding result is stored in the memory 86 as, for example, the decoding result FM1, and the sum of squares of the pixel unit difference from the decoding result FM0 is calculated by the calculation unit 87, This is realized by comparing this with a certain threshold TH0. That is, the sum of squares of the difference between frames is the threshold value.
When it is larger than TH0, it is considered that the convergence has not yet occurred, and the decoding operation is repeated. Therefore, it is necessary that some initial value is stored in the memory 86.

When the repetitive operation is continued, the operation unit 87
Copies the data of the decoding result FM0 stored in the memory 86 to the decoding result FM1. When the sum of squares of the inter-frame difference is less than or equal to the threshold value TH0, the calculation unit 87 regards the calculation as converged, ends the repetitive operation, and outputs the data of the calculation result FM0 to the outside as a decoded image.

The control unit 88 is composed of a so-called microcomputer, etc., controls the overall operation of the fractal decoding unit 42, and is provided with counters required for various processes (connections are omitted in the figure). .

Next, the decoded image generation adaptive filter 31 will be described.

In the decoded image generation filter 31, when the original image input as described above is coded by the high-speed fractal coding unit 32 and then further decoded by the fractal decoding unit 42 of the decoding unit 22. , Is converted into predicted image data that is predicted to be generated.

The decoded image generation adaptive filter 31 is used to reduce the amount of calculation at the time of encoding. That is, by analyzing that a decoded image can be decoded only up to a certain level of image quality with fractal coding of a certain block size, it is possible that the same level of image quality as that of the decoded image will be encoded and that the image quality will not change even if the image is further decoded. Known empirically. This is fractal coding, and
Since the high frequency components of the original image data are removed by fractal decoding, the failure rate of block matching is reduced, and as a result, even if the approximate block search is roughly performed, the S / N of the decoded image data and the original image data Is due to not decreasing.

Therefore, this decoded image generation adaptive filter 3
1 is a fractal encoding process of original image data, and
It is converted into predicted image data that is predicted to be decoded by the fractal decoding process (high-frequency components are removed).

The decoded image generation adaptive filter 31 is formed on the basis of a learning process between the decoded image data obtained by fractal encoding a plurality of original image data and further fractal decoding the original image data. It

FIG. 4 shows the configuration of the filter generation unit 101 (learning device for generating the coefficients used in the decoded image generation adaptive filter 31) which generates the decoded image generation adaptive filter 31 by the learning.

The fractal coding unit 111 and the fractal decoding unit 112 have the same basic configuration as the high-speed fractal coding unit 32 and the fractal decoding unit 42, and the description thereof will be omitted. The fractal coding unit 111 performs fractal coding on the supplied original image data and outputs the fractal decoding unit 112. Further, the fractal decoding unit 11
2 decodes the fractal-coded original image data,
The decoded image data is supplied to the learning unit 113.

The learning unit 113 generates a class code of each pixel by performing highly efficient compression encoding of the supplied original image data, for example, ADRC encoding processing. The learning unit 113 uses the original image data, the decoded image data, and the class code to calculate the optimum prediction coefficient for each class using the least square method or the like, and based on the calculated prediction coefficient, the decoded image generation adaptive filter. 31 is formed.

Here, the learning process will be described with reference to the flowchart of FIG.

In step S1, the fractal coding unit 111 performs fractal coding on the input original image data, and the fractal decoding unit 112 decodes the fractal coded original image data to obtain decoded image data. It is generated and supplied to the learning unit 113.

In step S2, the learning section 113
Learning data corresponding to the original image data and the decoded image data is formed, and for example, an array of (3 × 3) blocks shown in FIG. 6 around the pixels of the decoded image data corresponding to the pixels of the original image data is used as the learning data. use.

In step S3, the learning section 114
If it is determined whether the input of the original image data for one frame of the original image data has been completed, and if it is determined that the processing of, for example, one frame of the input original image data has not been completed, the process Proceeds to step S4.

In step 4, the learning section 113 executes class division processing of the input learning data. As described above, the pixel data of the original image data whose information amount is compressed by the ADRC process or the like is used as described above. In step S5, the learning unit 113 creates normal equations of equations (6) and (7) described later, and the process returns to step S1. That is, the class division process and the process of generating a normal equation associated with the process are repeated.

When it is determined in step 3 that the processing for one frame of the original image data has been completed, the processing proceeds to step S6. In step 6, the learning unit 113 solves the equation (8) described later using the matrix solution method to determine the prediction coefficient, and in step 7, stores the prediction coefficient in the decoded image generation adaptive filter 31 and performs the learning process. finish.

The process of step S5 (process of generating a normal equation) and the process of step S6 (process of determining a prediction coefficient) in FIG. 5 will be described in more detail. When the true value of the pixel of interest is y, its estimated value is y ', and the values of the surrounding pixels are x1 to xn, an n-tap linear linear combination with prediction coefficients w1 to wn for each class is given below. Formula (1)
Set as.

Y ′ = w1 · x1 + w2 · x2 + ... + wn · xn (1) In the equation (1), the prediction coefficient wi is an undetermined coefficient before learning.

As described above, learning is performed for each class,
When the number of data is m, according to the equation (1), yj ′ = w1 · xj1 + w2 · xj2 + ... + wn · xjn (2) (where j = 1, 2, ... m) When m> n, w1 to Since wn cannot be uniquely determined,
The elements of the error vector E are defined as ej = yj- (w1 * xj1 + w2 * xj2 + ... + wn * xjn) (3) (where j = 1, 2, ... The prediction coefficient that minimizes is obtained.

[0076]

[Equation 1] This is the so-called least squares method. Where equation (4)
The partial differential coefficient is obtained by the prediction coefficient wi of

[0077]

[Equation 2] Since each prediction coefficient wi may be determined so that the equation (5) is set to 0,

[Equation 3] And using the matrix

[Equation 4] Becomes This equation is generally called a normal equation. If this prediction equation wi is solved by using a general matrix solution method such as a sweeping method, the prediction coefficient wi is obtained, and the prediction coefficient wi is determined using the class code as an address.
Are stored in the decoded image generation adaptive filter 31, and the learning process is executed.

The decoded image generation adaptive filter 31 formed as described above, when converting the original image data into the transmission data, can be obtained without performing the fractal decoding again after the fractal coding by the conventional method. Since the image data can be obtained by the filter processing, the amount of calculation required for the processing is suppressed and the processing time is shortened. Further, since the decoded image generation adaptive filter 31 is formed by learning from a plurality of original image data, the original image data is encoded by the high-speed fractal encoding unit 32 and then decoded by the fractal decoding unit 42 again. It is possible to generate highly accurate (high S / N) predicted image data for the decoded image data at this time.

Next, referring to the flowchart of FIG.
The processing when the encoding unit 21 encodes the original image data into the transmission code will be described.

In step S11, the decoded image generation adaptive filter 31 filters the original image data to generate predicted image data, and the high-speed fractal coding unit 32,
The residual component adaptive quantizer 34 and the adder 33 are supplied.

In step S12, the adder 33
The prediction image data generated by the decoded image generation adaptive filter 31 is subtracted from the original image data to extract the residual component, which is output to the residual component adaptive quantization unit 34.

In step S13, the residual component adaptive quantization unit 34 executes the residual component adaptive quantization process.

Here, the residual component adaptive quantization processing by the residual component adaptive quantization unit 34 will be described with reference to the flowchart of FIG.

Incidentally, in explaining the residual component adaptive quantization processing, the range block and the domain block are defined as shown in FIGS. 9 to 11. That is, FIG.
As shown in, one image (for example, one frame) is divided into range blocks of size 8 × 8 shown in FIG. For example, by dividing an effective image of 720 pixels × 480 lines into range blocks,
A x60 range block is formed. It is assumed that the number Bij of each range block is defined by the number i that sequentially increases in the horizontal direction from the upper left corner of the image and the number j that sequentially increases in the vertical direction. Further, the pixels of each range block are assumed to be sequentially arranged in the right direction from upper left to p0, p1, p2 ... P62, and p63 as shown in FIG.

The domain block is, as shown in FIG.
The size shall be 16 × 16. It is assumed that the quantized data Qd obtained by ADRC encoding the pixel data of the domain block is linearly transformed, and the quantized data Qd ′ obtained by the linear transformation is moved in one pixel step within the search range.
The search range is defined as a range of (-8 to +7) in each of the horizontal and vertical directions as shown in FIG. 11 as an example. The position of this search range is k in the horizontal direction.
It is specified by the numbers (-8 to +7), and in the vertical direction, l
It shall be specified by a number (-8 to +7). Therefore, the domain block at the position within the search range is D
It will be specified in kl.

In step S21, the residual component adaptive quantization unit 34 obtains the maximum value MAX, the minimum value MIN, and the dynamic range DR of the pixel value for each block from the input predicted image data.

In step S22, the residual component adaptive quantizer 34 sets the block number counters i and j to i = j.
Initialize to = 0.

In step S23, the residual component adaptive quantizer 34 initializes the pixel number counter p to p = 0.

In step S24, the range block
Dynamic range D for residual component DF corresponding to Bij
R is read, and the residual component adaptive quantization unit 34 multiplies the coefficient α to obtain the threshold value TH2 (= DR × α).

In step S25, the residual component adaptive quantizer 34 obtains the dynamic range DR in the range block Bij, and determines the minimum value DR (min) of the dynamic range.
Less than, for example, range block
If the dynamic range DR in Bij is not smaller than the minimum value DR (min), that is, is equal to or larger than the minimum value DR (min), the process proceeds to step S26.

In step S26, the residual component adaptive quantization unit 34 determines whether or not the residual component DF of the pixel number p of the range block Bij is larger than the threshold value TH2, and the residual component adaptive quantizer 34 of the pixel number p of the range block Bij is determined. When it is determined that the residual component DF is not larger than the threshold TH2, that is, it is equal to or smaller than the threshold TH2, the process proceeds to step S27.

In step S27, the residual component adaptive quantization unit 34 sets a threshold value (− (residual component DF)) obtained by multiplying the residual component DF of the pixel number p of the range block Bij by a negative value.
If it is determined that it is not larger than TH2, that is, if it is smaller than the threshold TH2, the processing is
It proceeds to step S28.

In step S28, the residual component adaptive quantization unit 34 sets the quantization code of the residual component DF of the corresponding pixel p to 00. In step S29, the residual component adaptive quantization unit 34 sets the pixel number counter p to 63.
If it is determined that it is not within the range block, that is, if it is within the range block, the process Proceeds to step S30.

In step S30, the residual component adaptive quantization unit 34 increments the pixel number counter p by 1 (p = p + 1), and the process is step S27.
Then, the process is repeated after that.

In step S27, the range block
When it is determined that the residual component DF of the pixel number p of Bij is larger than the threshold value TH2, the process proceeds to step S31, and the residual component adaptive quantization unit 34 determines the residual component DF of the corresponding pixel p. The quantization code of is set to 01, and the process is
It proceeds to step S29.

In step S27, the range block
When it is determined that the negative residual component DF of the pixel number p of Bij is larger than the threshold TH2, the process proceeds to step S31, and the residual component adaptive quantization unit 34 determines the residual difference of the corresponding pixel p. The quantization code of the component DF is set to 10, and the process proceeds to step S28.

That is, until the quantization code corresponding to the pixels in the same range block is set, step S2
The processing from 6 to S32 is repeated.

When it is determined in step S29 that the pixel number counter p is larger than 63, that is, when it is determined that the pixel number is not within the range block, the process proceeds to step S33. In step S33, the residual component adaptive quantization unit 34 increments the block number counter i by 1 (i = i + 1).

In step S34, the residual component adaptive quantizing unit 34 determines whether or not the block number counter i is greater than 63, that is, as shown in FIG. If it is determined that it has not exceeded, the process proceeds to step S23.
Proceed to. That is, the processes of steps S23 to S34 are repeated until the process for one step is completed.

In step S34, if it is determined that the block number counter p is greater than 63, that is, if it is determined that the number for one stage in the frame is exceeded, as shown in FIG. Is step S35.
Proceed to.

In step S35, the residual component adaptive quantization unit 34 increments the block number counter j by 1 (j = j + 1). In step S36, the difference component adaptive quantization unit 34 determines whether or not the block number counter j is larger than 63, that is, as shown in FIG. 9, does not exceed the number of one column in the frame (( It is determined whether or not the processing for one frame is completed), and when it is determined that the processing has not been exceeded, the processing is step S2.
Go to 3. That is, the processing of steps S23 to S36 is repeated until the processing for one frame is completed.
If it is determined in step S36 that the block number counter j is larger than 63, that is, if it is determined that the processing for one frame has been completed, the processing ends.

If it is determined in step S25 that the dynamic range DR is smaller than the minimum value DR (min) of the dynamic range, in step S37
The residual component adaptive quantization unit 34 sets the quantization code of the residual component DF of the corresponding pixel p to 11, and the process proceeds to step S33.

Through the above processing, the residual component corresponding to each pixel is converted into a quantized code. That is, the dynamic range DR is obtained as the difference between the minimum value and the maximum value of the predicted pixel data, as shown in FIG. 12A. Therefore,
The residual component is adaptively quantized with respect to the dynamic range DR, so that it is accurately reproduced at the time of decoding. Therefore, in the process of step S24, as shown in FIG. 12B, the dynamic range DR is multiplied by a predetermined coefficient α to set the threshold TH2, and the quantized value is set according to the magnitude relationship with the threshold TH2.

In this example, as shown in FIGS. 12C and 13, when the residual component DF> threshold TH2 (step S
26 is Yes), the quantization value is set to 1, the quantization code is set to 01 (processing of step S31),-the residual component DF>
When the threshold value is TH2 (Yes in step S27), the quantization value is set to -1, and the quantization code is set to 10 (step S27).
32)), if threshold value TH2 ≧ residual component DF ≧ −threshold value TH2 (No in steps S26 and S27), the quantization value is set to 0 and the quantization code is set to 00.

Further, in step S25, the dynamic range DR in the range block Bij is obtained, and it is determined whether or not it is smaller than the minimum value DR (min) of the average values. For example, as shown in FIG. 14A, the dynamic range
Comparing the case where the DR is relatively large and the case where the dynamic range DR is relatively small as shown in FIG. 14B, it can be seen that the smaller the dynamic range DR, the smaller the difference between the predicted image data and the original image data. Therefore, the smaller the dynamic range DR, the more accurate the original image data will be generated without correcting the predicted image data by the residual component. Therefore, in the present example, when the minimum value DR (min) of the dynamic range DR is set as a predetermined value and the dynamic range DR in the range block is smaller than the minimum value DR (min) (in step S25, If Yes), it indicates that transfer of the residual component corresponding to the range block is unnecessary 11
Is the quantization code.

As shown in FIG. 15, when the residual component corresponding to each pixel in the range block is transferred, if the quantization code of the residual component at the beginning is 11, the subsequent 1 range block Do not transfer the quantization code of the residual component. By this processing, it is possible to avoid transferring the quantized code that does not affect the decoding, so that the transmission data can be made smaller, and the transfer speed and the processing speed can be improved.

Here, the description returns to the flowchart of FIG.

In step S14, the high-speed fractal coding unit 32 performs fractal coding processing on the input predicted image data.

Here, the fractal coding processing by the high-speed fractal coding unit 32 will be described with reference to the flowchart of FIG.

In step S51, the control unit 70 of the high speed fractal encoding unit 32 initializes the counters ij indicating the range block numbers with i = 0 and j = 0, respectively. In step S52, the control unit 70 sets the range block number counter i to 90 or more (i ≧).
90), that is, whether it is one stage or more of the range block shown in FIG. 9, and if it is determined that i ≧ 90 is not satisfied, the process proceeds to step S53.

In step S53, the range blocking unit 61 converts the search area of the input predicted image data into range blocks, and outputs the range blocks to the ADRC encoder 62. That is, 90 range blocking with j = 0 is executed first. In step S54, the ADRC encoder 62 performs the ADRC encoding process on the range block Bij input from the range blocking unit 61. That is,
In this case, the ADRC encoder 62 ADRC encodes the first range block B0000.

In step S55, the control unit 70
Initialize k and l, which are counters of domain block numbers. In this case, as shown in FIG. 11, counter k,
l is initialized to k = l = -8.

In step S56, the control unit 70
It is determined whether or not the counter k is 8 or more (k ≧ 8), that is, whether or not it is within the horizontal search range of the domain block shown in FIG. 11, for example, within the horizontal search range. When it is determined that, that is, k ≧ 80
If not, the process is step S5.
Proceed to 7.

In step S57, the search area cutout unit 65 cuts out the domain block, and at the same time, the domain blocking unit 66 executes the domain blocking process, and outputs the domain-blocked predicted image data to the ADRC encoder 67. To do. In this case, the domain blocking unit 66 forms the domain block D-8-8 by the first domain blocking process, and outputs it to the ADRC encoder 67.

In step S 58, the ADRC encoder 67 ADRC-encodes the domain block Dkl and outputs it to the 1/4 reduction unit 68. In step S59, the 1/4 reduction unit 68 reduces the horizontal and vertical block sizes of the quantized data Qd input by the ADRC encoder 67 to halves, and reduces the quantized data Qd to a size of 8 × 8, that is, a size of 1/4. . This reduced domain block is represented as Dkl '.

In step S60, the rotating mirror image processing unit 69 sets the rotation angle R of the rotating operation to 0. That is, the conversion operation of only reduction is performed on the quantized data Qd of the domain block. In step S61, the rotating mirror image processing unit determines whether or not the rotation angle is R ≧ 360 °, that is, whether or not the rotation has completed one round, and R ≧ 3.
If it is determined that the number is not 60, that is, one round has not been completed, the process proceeds to step S62.

In step S62, the rotating mirror image processing unit 69 rotates the domain block Dkl ′ by the angle R. In this case, when the rotation angle R = 0, it means that the domain block Dkl ′ is not rotated.

In step S63, the rotating mirror image processing unit 69 obtains the sum of squares of the differences between the pixels of the rotated domain block Dkl 'and the range block Bij, and uses this as an evaluation value in the comparison unit 63 as an evaluation value table. Store. In the case of R = 0, the evaluation value between the domain block and the range block on which the reduction operation has been performed is calculated and stored in the table.

In step S64, the rotating mirror image processing unit 69 horizontally flips the rotated domain block Dkl '. Hereinafter, this rotated and inverted domain block will be referred to as Dkl ″.

In step S65, the rotating mirror image processing unit 69 calculates the sum of squares of the differences between the pixels of the domain block Dkl ″ and the domain range block Bij that have been rotated and inverted, and outputs the sum to the comparing unit 63 for evaluation. The value is stored in the evaluation value table. In the case of R = 0, the evaluation value between the domain block and the range block on which the reduction and inversion operations have been performed is calculated and stored in the table.

In step S66, the rotating mirror image processing unit 69 adds 90 ° to the rotation angle R (+ 90 °), and the process returns to step S61. That is, the processes of steps S61 to S66 are repeated until the rotation process is performed once (R ≧ 360 ° in step S61), and the obtained evaluation value is stored in the evaluation value table of the comparison unit 63. Is stored. As a result, the same processing as described above is performed for each of the rotation angles R of R = 0, 90 °, 180 °, and 270 °.

As an example, for the domain block D-8-8, the evaluation values for the eight types of conversion parameters described above are obtained. That is, the first parameter (1/2 reduction), the second parameter (1/2 reduction and horizontal inversion), the third parameter (1/2 reduction and 90
Rotation, 4th parameter (1/2 reduction and 180
Rotation, 5th parameter (1/2 reduction and 270
° rotation), 6th parameter (1/2 reduction, 90 ° rotation and horizontal flip), 7th parameter (1/2 reduction, 1
Evaluation values for 80 ° rotation and left / right inversion) and the eighth parameter (1/2 reduction, 270 ° rotation and left / right inversion) are obtained, respectively.

Therefore, in step S61, the evaluation value for each of the first to eighth parameters is obtained for one domain block Dkl until it is determined that the rotation angle R ≧ 360. Become. If it is determined in step S61 that the rotation angle is R ≧ 360, that is, if the domain block Dkl has made one round, the value of the counter k is incremented by 1 in step S67. In other words, the process of step S67 causes the domain block to be searched within the search range.
The position of Dkl is horizontally shifted by one pixel. And
The process returns to step S56, and as described above,
Reduction about domain blocks in shifted positions,
Rotation and left-right inversion operations are performed to obtain evaluation values for eight conversion parameters. This evaluation value is also stored in the evaluation value table.

If the position of the domain block is horizontally shifted by one pixel in the search range and it is determined in step S56 that k ≧ 8, step S6 is performed.
In 8, whether the counter l is 8 or more (l ≧ 8),
That is, it is determined whether or not the lower limit of the search range of the domain block has been reached. When it is determined in step S68 that the counter l is not 8 or more, that is, the domain block has not reached the lower limit value,
In step S69, the value of l is incremented by 1, the counter k is initialized to k = -8, and the process returns to step S57 (domain block).

That is, by the processing of step S69,
The vertical position of the domain block in the search range is 1
By shifting to the lower side of the line and incrementing the value of k on the line, the position of the domain block is shifted in the horizontal direction, and the evaluation value is calculated at each position.

In step S68, l ≧ 8,
That is, when it is determined that the domain block has reached the lower limit value of the search range, the process proceeds to step S70. In step S70, the minimum value determination unit 64
The minimum value of the plurality of evaluation values (sum of squared differences as described above) stored in the evaluation value table stored in the comparison unit 63 is detected and output to the transmission format encoding unit 35.

In step S71, the transmission format coding unit 35 creates coded data corresponding to the detected minimum value, and generates the dynamic range DR and minimum value MIN of the domain block and the minimum evaluation value. The position of the domain block (values of k and l), the parameter, and the adaptive quantized residual component are synthesized.
Convert to transmission data that is compatible with the transmission format.

In step S72, the transmission format encoder 35 outputs the transmission data. Step S
In 73, the control unit 70 increments the counter i by 1, and the process returns to step S52. That is, by incrementing the counter i by 1, the encoding process for the next range block is
It starts from step S52.

If it is determined in step S52 that i ≧ 90, that is, if the processing of the range block for one stage shown in FIG. 9 is completed, the counter j is 60 or more (j ≧ 60) in step S74. Or not,
That is, it is determined whether or not the range blocks of all stages have been encoded in the vertical direction, and for example, the counter j is j ≧ j.
If not 60, that is, if it is determined that there is a stage that has not been encoded yet in the vertical direction, the process proceeds to step S75.

At step S75, the control section 70
The counter j is incremented by 1 and the counter i is incremented by i
Is initialized to = 0, the process returns to step S53, and the subsequent processes are repeated.

At step S74, the counter j is set to j.
If ≧ 60, that is, if it is determined that all pixels for one frame have been encoded, the processing of all range blocks of one frame ends.

Here, the description returns to the flowchart of FIG.

Through the above processing, the original image data is converted into transmission data and output.

Next, with reference to the flowchart of FIG. 17, the processing of the decoding unit 22 when decoding the transmission data generated by the processing described with reference to the flowchart of FIG. 7 to generate decoded image data explain.

In step S91, the transmission parameter separating section 41 separates the fractal-coded image data and the adaptively quantized residual component from the coded image data to obtain the fractal-coded image data. The residual component output to the fractal decoding unit 42 and adaptively quantized is output to the residual component adaptive inverse quantization unit 43.

In step S92, the fractal decoding unit 42 executes the fractal decoding process of the fractal-coded image data input from the transmission parameter separation unit 41.

Here, the fractal decoding processing by the fractal decoding unit 42 will be described with reference to the flowchart in FIG.

In step S111, the control unit 88
Controls the memory 86 to initialize the calculation results FM0 and FM1. In step S112, the control unit 88 initializes the range block number counters i and j by setting them to 0.

In step S113, the domain block cutout unit 81 cuts out the domain block according to the position information of the domain block in the received data, and outputs it to the ADRC encoder 82. Here, the domain block cut out corresponding to the range block Bij to be decoded is represented as Dij.

In step S114, the ADRC encoder 82 ADRC-encodes this domain block Dij,
Output to the 1/4 reduction unit 83. In step S115, the 1/4 reduction unit 83 reduces only the quantized data in the encoded data generated by the ADRC encoder and outputs it to the rotating mirror image processing unit 84.

In step S116, the rotating mirror image processing unit 84 inputs the reduced domain block Dij.
Rotate the quantized data of. Further, step S11
In 7, the rotating mirror image processing unit 84 inverts (mirror image processes) the quantized data of the rotated domain block Dij, and outputs it to the ADRC decoder 85.

In step S118, the ADRC decoder 85 has received the data of the domain block after the conversion operation, executes the ADRC decoding process using the conversion parameter, and decodes the decoded data of the range block Bij, that is, the decoded data. Predicted image data can be obtained. Step S
At 119, the control unit 88 outputs the decoded predicted image data to the memory 86 and stores it as the calculation result FM0 at the position of the range block Bij.

In step S120, the control unit 88
Is whether the counter i is 90 or more (i ≧ 90),
That is, 1 set in the horizontal direction as shown in FIG.
It is determined whether or not the processing of the range block for one step is completed, and the counter i is not 90 or more (i ≧ 90), that is, the processing of the range block for one step set in the horizontal direction is completed. If it is determined that there is not, step S1
In 21, the control unit 88 increments the value of the counter i by 1, and the process returns to step S113. That is, the processing of steps S113 to S121 is repeated until the processing of the range block for one stage set in the horizontal direction is completed.

When it is determined in step S120 that the counter i is 90 or more (i ≧ 90), that is, the processing of the range blocks of all the stages set in the vertical direction is completed, the processing is performed in step S120. It proceeds to S122.

In step S122, the control unit 88.
Is whether or not the counter j is 60 or more (j ≧ 60),
That is, it is determined whether the processing of the range block in the vertical direction is completed (is the processing of one frame is completed),
For example, the counter j is not 60 or more (j ≧ 60),
That is, when it is determined that the processing of the range block in the vertical direction is not completed, the processing is step S12.
Go to 3.

In step S123, the control unit 88
Increments the counter j by 1 (j = j +
1), the counter i is set to 0, and the process proceeds to step S113. That is, the processing of steps S113 to S123 is repeated until the processing for one frame is completed.

If it is determined in step S122 that the counter j is 60 or more (j ≧ 60), that is, if the processing of the range block in the vertical direction has been completed, the processing proceeds to step S124.

In step S124, the calculation unit 87
Calculates the sum of squares S of the differences between the decoded predicted image data of the calculation result FM0 stored in the memory 86 and the decoded predicted image data of the calculation result FM1.

In step S125, the calculation unit 87
Determines whether the sum of squares S is larger than a predetermined threshold TH3 indicating the convergence of the calculation, that is, whether the calculation is converged, and the sum of squares S is a predetermined value indicating the convergence of the calculation. Threshold TH3
When it is determined that it is not larger, that is, the calculation has not converged, the process proceeds to step S126. In step S126, the calculation unit 87 sets the decoded prediction image data of the calculation result FM0 stored in the memory 86 to F
Copy to M1, the process returns to step S112,
The subsequent processing is repeated.

At step S125, the sum of squares S is
When it is determined that the calculation is larger than the predetermined threshold TH3 indicating the convergence of the calculation, that is, the calculation has converged, the process proceeds to step S127.

In step S127, the calculation unit 87
Outputs the calculation result FM0 stored in the memory 86 to the adder 44 as decoded predicted image data.

Through the above processing, the fractal-coded predicted image data is fractal-decoded.

Here, the description returns to the flowchart of FIG.

In step S93, the fractal decoding unit 42 outputs the dynamic range DR of the decoded prediction image data in units of blocks and the minimum value MIN to the residual component adaptive dequantization unit 43.

In step S94, the residual component adaptive dequantization unit 43 executes the residual component adaptive dequantization process,
The dequantized residual component and the pixel value of the decoded predicted image data are added to generate decoded image data, and the process ends.

Now, the residual component adaptive dequantization process by the residual component adaptive dequantization unit 43, the adder 44, and the clipping processing unit 45 will be described with reference to the flowchart in FIG.

In step S141, the residual component adaptive dequantization unit 43 initializes the counters i and j indicating the range block numbers (i = j = 0).

In step S142, the residual component adaptive dequantization unit 43 initializes the counter p indicating the pixel number (p = 0).

In step S143, the residual component adaptive dequantization unit 43 determines whether the first quantization code of the range block Bij is 11, that is, as shown in FIG. That is, since the residual component of the range block Bij has not been transferred, the determination is made. For example, the quantization code is not 11, that is, each range block Bij If it is determined that the quantized code corresponding to the pixel has been transferred, the process proceeds to step S
Proceed to 144.

In step S 144, the residual component adaptive dequantization unit 43 acquires the dynamic range DR and the minimum value MIN of the range block Bij input from the fractal decoding unit 42.

In step S145, the residual component adaptive dequantization unit 43 determines whether the quantization code of the residual component of the pixel number p of the range block Bij is 01,
For example, if it is determined that the value is not 01, the process is
It proceeds to step S146.

In step S146, the residual component adaptive inverse quantization unit 43 determines whether or not the quantization code of the residual component of the pixel number p of the range block Bij is 10.
For example, if it is determined that the value is not 10, the process is
It proceeds to step S147.

In step S147, the residual component adaptive dequantization unit 43 considers that the quantization code of the pixel number p is 00, outputs the corresponding residual component as 0, and the adder 44 responds. The pixel value of the decoded image data is added to the pixel value of the decoded predictive image data, and the pixel value of the decoded image data is output to the clipping processing unit 45.

In step S148, the clipping processing unit 45 confirms whether or not the decoded image data pixel value is within a range that can be displayed as an image. If the decoded image data pixel value is beyond the range, the clipping pixel value is set. Clipped to the maximum range (replace the calculated pixel value with the maximum range pixel value) and output as decoded image data.

In step S149, the residual component adaptive dequantization unit 43 determines whether or not the residual components corresponding to all the pixels in the range block Bij have been dequantized, that is, the pixel number counter p is 64. It is determined whether or not (p ≧ 64), and for example, the residual components corresponding to all the pixels in the range block Bij are not dequantized,
That is, the pixel number counter p is 64 or more (p ≧ 6
If it is determined not to be 4), the process proceeds to step S150.

In step S150, the residual component adaptive dequantization unit 43 increments the pixel number counter p by 1, and the process returns to step S145.

When it is determined in step S145 that the quantization code of the residual component of the pixel number p of the range block Bij is 01, the residual component adaptive dequantization unit 43 determines in step S151 The residual component of the dynamic range DR multiplied by a predetermined coefficient β,
The value obtained by multiplying the quantized value corresponding to the quantized code is output as the residual component to the adder 44, and the adder 44 adds this to the pixel value of the decoded predicted image data and the clipping processing unit 45. To step S151.
Proceed to. That is, since the quantized code 01 has a quantized value of 1 as shown in FIG. 13, the residual component becomes DR × β, and the value is added to the pixel value of the decoded predicted image data. become.

When it is determined in step S146 that the quantization code of the residual component of the pixel number p of the range block Bij is 10, the residual component adaptive dequantization unit 43 determines in step S152 that the corresponding pixel has the corresponding pixel. The residual component of the dynamic range DR multiplied by a predetermined coefficient β,
The value obtained by multiplying the quantized value corresponding to the quantized code is output as the residual component to the adder 44, and the adder 44 adds this to the pixel value of the decoded predicted image data and the clipping processing unit 45. To step S151.
Proceed to. That is, since the quantization code 10 has a quantized value of −1 as shown in FIG. 13, the residual component is −DR ×
β, which is added to the pixel value of the decoded predicted image data.

In step S149, it is determined that the residual components corresponding to all the pixels in the range block Bij have not been dequantized, that is, the pixel number counter p is not 64 or more (p ≧ 64). In this case, the residual component adaptive dequantization unit 43 uses the counter i of the range block Bij.
Is incremented by 1, and in step S154, whether i is larger than 89 (i> 89), that is, as shown in FIG. 9, the inverse quantization process is performed on all range blocks for one stage in the horizontal direction. If it is determined whether or not it has been executed and, for example, i> 89, that is, it is determined that the dequantization processing has not been executed in all range blocks for one stage in the horizontal direction, the processing proceeds to step The process returns to S142 and the subsequent processes are repeated.

In step S154, 1 in the horizontal direction
If it is determined that the inverse quantization process has not been executed in all range blocks of the stage, the process proceeds to step S155. In step S155, the residual component adaptive inverse quantization unit 43 increments the counter j of the range block Bij by 1, and in step S156, whether j is larger than 59 (i> 59), that is, in FIG. As shown, it is determined whether or not the inverse quantization processing has been executed in the range blocks (range blocks for one frame) at all stages in the vertical direction. For example, i> 5
If it is 9, that is, if it is determined that the dequantization processing is not executed in the range blocks of all the stages in the vertical direction, the processing returns to step S142, and the processing thereafter is repeated. In step S156, if i> 59, that is, if it is determined that the inverse quantization process has been executed in the range blocks of all the stages in the vertical direction, the process ends.

In step S143, the first quantization code of the range block Bij is 11, that is,
It is determined that the residual component of the range block Bij has not been transferred, and in step S157, the residual component adaptive inverse quantization unit 43 outputs the residual component corresponding to all the pixels of the block number Bij as 0. Then, the adder 44 outputs the pixel value of the corresponding decoded predicted image data to the clipping processing unit 45 as it is.

In step S158, the clipping processing section 45 performs clipping processing on all the pixel values of the range block Bij and outputs it as decoded image data.

Steps S147, S151, S152,
More specifically, the processing of S157 and S157, as shown in FIG. 20, shows that the dynamic range DR and the predetermined coefficient β with respect to the quantized value (empirically β = 1/2 is a reasonable value. (Known) and inversely quantized by multiplying the residual component adaptively to the dynamic range DR, and by adding the decoded predicted image data to this residual component, the decoded image data It can be reproduced accurately.

Also, as shown in FIG. 13, a 2-bit AD
In the RC processing, when the dynamic range DR of the range block is smaller than a predetermined minimum value, it is not transferred to the decoding unit 22, so the amount of calculation related to decoding can be suppressed, and the processing speed can be increased. It becomes possible to do.

In the above processing, the case where the quantization code in the ADRC processing is 2 bits has been described, but the present invention is not limited to this, and the quantization code may be set with other numbers of bits. Further, the setting of the range block and the domain block is not limited to the setting as shown in FIGS. 9 to 11, and may be a range block having a pixel number other than this and a domain block, or a range block having a number other than this. , And the number of domain blocks.

In the above example, the case where the dynamic range DR is multiplied by a predetermined coefficient α to set a quantized value in the residual component adaptive quantization processing has been described. It does not necessarily have to be a constant value, and may be a value that is adaptively set according to the S / N of the original image data and the predicted image data. Figure 21
Shows the configuration of the encoding unit 21 that can adaptively obtain the coefficient α. At this time, the residual component adaptive quantization unit 34 supplies the original image data, the residual component, and
The predicted image data will be input. FIG. 22 shows the coefficient α provided in the residual component adaptive quantization unit 34 at this time.
3 shows the configuration of the calculation unit of.

The α value variable quantization unit 131 changes the α value within a range of possible values, performs quantization by ADRC processing with DR × α as a threshold value as described above, and inversely quantizes the quantization code. It is output to the unit 132. The dequantization unit 132 dequantizes the quantization code of the residual component input from the α variable quantization unit 131 by multiplying the quantization value corresponding to the quantization code by the coefficient β to generate the residual component. , And output to the adder 133. The adder 133 adds the dequantized residual component and the predicted image data to generate decoded image data, and outputs the decoded image data to the S / N calculation unit 134. The S / N calculation unit 134 calculates the S / N of the original image data and the decoded image data, and outputs the S / N to the S / N evaluation unit 135. The S / N evaluation unit 135 stores the S / N input from the S / N calculation unit 134 in association with the coefficient α, and when all the coefficients α have been calculated, the S / N becomes the maximum coefficient. Select α as the coefficient used for quantization.

Next, the calculation unit for the coefficient α shown in FIG.
The process for obtaining the coefficient α that maximizes N will be described.

In step S191, the variable α quantizer 131 initializes the coefficient α. That is, the α-value variable quantization unit 131 sets the coefficient α to 0 when the coefficient α is set to the smallest possible value, for example, when 0 ≦ α ≦ 1 (in this case).

In step S192, the α value variable quantization unit 131 quantizes each residual component using the value obtained by multiplying the dynamic range DR by the coefficient α as a threshold value, and the quantized residual component is dequantized by the inverse quantization unit 132. Output to. The quantization processing is, for example, the method shown in FIG. 12, and the description thereof will be omitted.

In step S193, the inverse quantizer 1
32 multiplies the quantized value quantized in the process of step S192 by the dynamic range DR and a predetermined coefficient β to dequantize the residual component, and adds the dequantized residual component to the adder 1
To 33. In step S194, the adder 1
33 adds the dequantized residual component and the predicted image data to generate decoded image data, and outputs the decoded image data to the S / N calculation unit 134.

In step S195, the S / N calculator 1
34 is based on the input original image data and decoded image data
The S / N is calculated and output to the S / N evaluation unit 135. Step S
In 196, the S / N evaluation unit 135 determines that the current coefficient α
The S / N is stored in association with the value of. In step S197, the S / N evaluation unit 135 determines whether or not the S / N has been obtained for all possible values of the coefficient α, and, for example,
If it is determined that S / N has not been obtained for all the coefficients α, the process proceeds to step S198.

In step S198, the variable α quantizer 131 increments the value of the coefficient α by a predetermined step width, the process returns to step S192, and the subsequent processes are repeated. That is, the processes of steps S192 to S198 are repeated until the S / N is obtained for all the possible coefficients α.

If it is determined in step S197 that S / N has been obtained for all possible values of the coefficient α, in step S199, the S / N evaluation section 135
Sets the coefficient α for which the calculated S / N has the maximum value as the coefficient used in the residual component adaptive quantization processing.

By using the coefficient α set by the above processing, the residual component can be quantized by the optimum coefficient α of S / N, so that the original image data can be reproduced more faithfully. It can be reproduced as

Further, the accuracy of the inverse quantization can be improved by adaptively setting the inverse quantization in the residual component adaptive inverse quantization unit 43 according to the dynamic range DR. That is, as shown in FIG. 24, it is known that the residual component generated by the fractal coding process is biased in the existence distribution due to the dynamic range DR. Therefore, the dynamic range from multiple image data
Dynamic range is calculated by calculating the distribution of the existence of DR and residual components.
Inverse quantization may be adaptively performed according to the existence distribution of DR and the residual component. For example, dynamic range DR = 3
The peak value of the residual distribution of the residual component at 0 is, for example,
In the case of +20, the method is such that dequantization is performed to +20 when the dynamic range DR of decoded predicted image data DR = 30 is also used in dequantization. By such processing, a value having a high existence frequency can be obtained by inverse quantization, so that decoded image data can be generated with higher accuracy.

Further, as shown in FIG. 25, DRC (Digit
An al Reality Creation) processing unit 141 and a memory 142 storing a fractal image coefficient set may be further provided. In this case, the DRC processing unit 141 reads the fractal image coefficient set from the memory 142, and determines the pixel value to be interpolated when, for example, improving the horizontal resolution of the decoded image data generated by the high quality image decoding unit 22. The calculation can be performed using the coefficient set for the fractal image to improve the resolution, and the image quality of the decoded image data can be improved.

In the above example, the encoding process,
And, as a decoding method, fractal coding processing,
Also, the fractal decoding process has been described as an example, but the encoding and decoding process methods may be other than this, for example, JPEG (Joint Photographic
Encoding process using Experts Group), etc., and
It may be decryption processing.

The series of processes described above can be executed by hardware, but can also be executed by software. When a series of processes is executed by software, various functions can be executed by installing a computer in which a program configuring the software is incorporated in dedicated hardware or various programs. It is installed from a recording medium into a possible general-purpose personal computer or the like.

26 and 27 show the configuration of an embodiment of a personal computer in which the encoding unit 21 and the decoding unit 22 are realized by software. The CPUs 201 and 301 of the personal computer are
Controls the overall operation of the personal computer. In addition, the CPUs 201 and 301 are input units 206 and 30 including a keyboard and a mouse from a user via the buses 204 and 304 and the input / output interfaces 205 and 305.
When a command is input from 6, the ROM (Read O
nly Memory) 202, 302 to execute the program. Alternatively, the CPUs 201 and 301 are
Magnetic disk 21 connected to drives 210 and 310
1, 311, optical disks 212, 312, magneto-optical disks 213, 313, or semiconductor memories 214, 314.
The program read from the memory and installed in the storage units 208 and 308 is stored in a RAM (Random Access Memory) 20.
Load to 3,303 and execute. As a result, the functions of the image processing apparatus described above are realized by software. Further, the CPUs 201 and 301 are connected to the communication unit 20.
It controls 9, 309 to communicate with the outside and exchange data.

The recording medium on which the program is recorded is
As shown in FIGS. 26 and 27, separately from the computer,
A magnetic disk 211, 311 on which the program is recorded, which is distributed to provide the program to the user.
(Including a flexible disk), an optical disk 212,
312 (CD-ROM (Compact Disc-Read Only Memory), DVD
(Including Digital Versatile Disc), magneto-optical disc 213, 313 (including MD (Mini-Disc)), semiconductor memory 214, 314, etc. The ROMs 202 and 302 in which the programs are recorded and the storage unit 20 that are provided to the user in the closed state
The hard disk included in 8, 308 and the like.

In the present specification, the steps for writing the program recorded on the recording medium are not limited to the processes performed in time series in the order described, but not necessarily in time series. , Which include processes executed in parallel or individually.

Further, in this specification, the system means
It represents the entire apparatus composed of a plurality of devices.

[0194]

According to the first information processing apparatus and method and the program of the present invention, from the input image data,
After being encoded, it generates predicted image data that is predicted to be decoded, encodes the predicted image data, calculates difference data between the input image data and the predicted image data, and predicts the calculated difference data. Prediction image data that is adaptively quantized and encoded corresponding to the feature amount of image data, and difference data that is adaptively quantized with respect to the feature amount of the predicted image data are output as encoded data. .

A second information processing apparatus and method of the present invention,
According to the program, when the original image data is encoded and then further decoded, the predicted image data predicted to be decoded is encoded image data,
The coded image data and the quantized difference data are separated from the coded data consisting of the difference data between the original image data and the predicted image data, which are adaptively quantized with respect to the feature amount of the predicted image data. The encoded image data is decoded into predicted image data, the quantized difference data is adaptively dequantized corresponding to the feature amount of the predicted image data, and the predicted image data and the dequantized difference are obtained. The decoded image data is generated based on the data, and the generated decoded image data is output.

In any case, as a result, it is possible to realize high-speed encoding and decoding of image data, and it is possible to decode encoded image data with high image quality.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an embodiment of an image transfer system to which the present invention has been applied.

2 is a block diagram showing a configuration of a high-speed fractal coding unit in FIG.

3 is a block diagram showing a configuration of a fractal decoding unit in FIG.

FIG. 4 is a block diagram illustrating a filter generation unit.

FIG. 5 is a flowchart illustrating a learning process for generating a decoded image generation adaptive filter.

FIG. 6 is a diagram illustrating a learning process for generating a decoded image generation adaptive filter.

FIG. 7 is a flowchart illustrating an encoding process.

FIG. 8 is a flowchart illustrating residual component adaptive quantization processing.

FIG. 9 is a diagram illustrating a range block and a domain block.

FIG. 10 is a diagram illustrating a range block.

FIG. 11 is a diagram illustrating a search range of a domain block.

FIG. 12 is a diagram illustrating ADRC processing.

FIG. 13 is a diagram illustrating a relationship between a quantization code of ADRC and a quantization value of a residual component.

FIG. 14 is a diagram illustrating ADRC processing.

FIG. 15 is a diagram illustrating ADRC processing.

FIG. 16 is a flowchart illustrating a fractal encoding process.

FIG. 17 is a flowchart illustrating a decoding process.

FIG. 18 is a flowchart illustrating a fractal decoding process.

FIG. 19 is a flowchart illustrating residual component adaptive quantization processing.

FIG. 20 is a flowchart illustrating residual component adaptive quantization processing.

FIG. 21 is a diagram illustrating a configuration for setting an optimum value of coefficient α.

FIG. 22 is a diagram illustrating a configuration for setting an optimum value of coefficient α.

FIG. 23 is a flowchart illustrating a process of setting an optimum value of coefficient α.

FIG. 24 is a diagram illustrating an inverse quantization process corresponding to a dynamic range.

FIG. 25 is a diagram illustrating a configuration in which a DRC processing unit is provided in the decoding unit 22.

FIG. 26 is a diagram illustrating a medium.

FIG. 27 is a diagram illustrating a medium.

[Explanation of symbols]

11 image transfer system, 21 coding unit, 22 decoding unit, 31 decoded image generation adaptive filter, 32 high-speed fractal coding unit, 33 adder, 34 residual component adaptive quantization unit, 35 transmission format coding unit, 41 transmission Parameter separation unit, 42 Residual component adaptive quantization unit, 43
Residual component adaptive dequantization unit, 44 adder, 45 clipping processing unit, 61 range blocking unit, 62 AD
RC encoder, 63 comparison unit, 64 minimum value determination unit, 6
5 Search area cutout unit, 66 domain blocking unit, 67 ADRC encoder, 68 1/4 reduction unit, 69
Rotating mirror image processing unit, 70 control unit, 81 domain block cutout unit, 82 ADRC encoder, 83 1/4 reduction unit, 84 rotating mirror image processing unit, 85 ADRC encoder, 8
6 memory, 87 operation part, 88 control part, 101 filter generation part, 111 fractal coding part, 112
Fractal decoding unit, 113 learning unit, 131 α-valued variable quantization unit, 132 inverse quantization unit, 133 adder, 13
4 S / N calculation unit, 135 S / N evaluation unit, 141 DRC processing unit, 142 memory

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Kazushi Yoshikawa             6-735 Kita-Shinagawa, Shinagawa-ku, Tokyo Soni             -Inside the corporation (72) Inventor Hideo Kasama             6-735 Kita-Shinagawa, Shinagawa-ku, Tokyo Soni             -Inside the corporation (72) Inventor Tsuguhiko Haga             6-735 Kita-Shinagawa, Shinagawa-ku, Tokyo Soni             -Inside the corporation F term (reference) 5C059 KK11 KK23 MA28 MA43 MD02                       RA01 SS20 TA52 TA69 TB08                       TC02 TC10 TD11 UA02 UA05                       UA11 UA39                 5C078 AA04 BA44 CA31 DA01                 5J064 AA01 BB01 BB03 BB12 BC01                       BC11 BC26

Claims (18)

[Claims] 1. An information processing apparatus for outputting coded data based on input image data, wherein the generation means generates predicted image data that is predicted to be decoded after being coded, from the input image data. An encoding unit that encodes the predicted image data, a difference calculation unit that calculates difference data between the input image data and the predicted image data, and the difference data calculated by the difference calculation unit,
Quantization means for adaptively quantizing in accordance with the feature amount of the predicted image data, the predicted image data encoded by the encoding means, and adaptively for the feature amount of the predicted image data An information processing apparatus comprising: an output unit that outputs the quantized difference data as encoded data. 2. The information processing apparatus according to claim 1, wherein the encoding means encodes the predicted image data by performing a fractal encoding process. 3. The quantizing means, for each block,
The quantization is performed according to the value of the dynamic range so as to indicate to the output means that output of the residual component of the block as encoded data is unnecessary. The information processing device described. 4. The quantizing means, for each block,
According to the value of the dynamic range, when the value of the dynamic range is smaller than a predetermined threshold value, it indicates to the output means that the output of the residual component of the block as encoded data is unnecessary. The information processing apparatus according to claim 3, wherein the information processing apparatus is quantized into. 5. The information processing apparatus according to claim 1, wherein the feature amount is a dynamic range of pixel values for each block of the predicted image data. 6. The quantizing means, for each block,
Comparing the pixel value of the predicted image data with a value obtained by multiplying the dynamic range by a predetermined coefficient, and adaptively quantizing the difference data calculated by the difference calculating means in accordance with the comparison result. The information processing apparatus according to claim 5, wherein: 7. The predetermined coefficient is set for each block of the predicted image data in correspondence with the dynamic range,
7. The information processing apparatus according to claim 6, further comprising coefficient adaptive calculation means for adaptively calculating. 8. An information processing method of an information processing apparatus for outputting coded data based on input image data, wherein predicted image data predicted to be decoded after being encoded is input from the input image data. A generating step of generating, an encoding step of encoding the predicted image data, a difference calculation step of calculating difference data between the input image data and the predicted image data, and a calculation of the difference calculation step Quantization step of adaptively quantizing the difference data corresponding to the feature amount of the prediction image data, the prediction image data encoded in the process of the encoding step, and features of the prediction image data An output step of outputting the difference data adaptively quantized with respect to the quantity as encoded data. 9. A program for controlling an information processing apparatus that outputs encoded data based on input image data, the predictive image being predicted to be decoded after being encoded from the input image data. A generation control step of controlling generation of data, an encoding control step of controlling encoding of the predicted image data, a difference calculation control step of controlling calculation of difference data between the input image data and the predicted image data, A quantization control step of controlling an adaptive quantization of the difference data, the calculation of which has been controlled in the processing of the difference calculation control step, corresponding to the feature amount of the predicted image data; and a processing of the coding control step. Of the prediction image data whose encoding is controlled by, and the encoding data of the difference data adaptively quantized with respect to the feature amount of the prediction image data. Recording medium from which a computer readable program is recorded, characterized in that it comprises an output control step of controlling the output of the data. 10. A computer for controlling an information processing apparatus that outputs encoded data based on input image data, is provided with predictive image data that is predicted to be decoded after being encoded from the input image data. A generation control step of controlling generation, a coding control step of controlling coding of the predicted image data, a difference calculation control step of controlling calculation of difference data between the input image data and the predicted image data, A quantization control step of controlling adaptive quantization of the difference data, the calculation of which is controlled in the processing of the difference calculation control step, corresponding to the feature amount of the predicted image data, and a code in the processing of the coding control step. Encoding of the difference data that is adaptively quantized with respect to the prediction image data whose encoding is controlled and the feature amount of the prediction image data. Program for executing an output control step of controlling the output of the. 11. An information processing apparatus for decoding encoded data of original image data into decoded image data corresponding to the original image data, when the original image data is further decoded after being encoded,
Coded image data obtained by coding the predicted image data predicted to be decoded, and the difference between the original image data adaptively quantized with respect to the feature amount of the predicted image data and the predicted image data From encoded data consisting of data,
Separation means for separating the coded image data and the quantized difference data, a decoding means for decoding the coded image data separated by the separation means into the predicted image data, and the quantized difference Dequantizing means for adaptively dequantizing data corresponding to the feature amount of the predicted image data, the predicted image data, and the decoded image data based on the dequantized difference data. An information processing apparatus, comprising: a generation unit that generates a. And an output unit that outputs the decoded image data generated by the generation unit. 12. The information processing apparatus according to claim 11, wherein the decoding unit decodes the predicted image data by performing a fractal decoding process on the encoded image data. 13. The information processing apparatus according to claim 11, wherein the characteristic amount is a dynamic range of pixel values for each block of the predicted image data. 14. The inverse quantization means adaptively inverses the dynamic range for each block by multiplying the dynamic range by a predetermined coefficient and further by multiplying the quantized difference data by the multiplied value. 14. Quantization, characterized in that
The information processing device according to 1. 15. The coefficient adaptive calculation means for adaptively calculating the predetermined coefficient for each block of the predicted image data according to the dynamic range, according to claim 14. Information processing equipment. 16. An information processing method of an information processing apparatus for decoding encoded data of original image data into decoded image data corresponding to the original image data, when the original image data is further decoded after being encoded. ,
Coded image data obtained by coding the predicted image data predicted to be decoded, and the difference between the original image data adaptively quantized with respect to the feature amount of the predicted image data and the predicted image data From encoded data consisting of data,
A separation step of separating the coded image data and the quantized difference data; a decoding step of decoding the coded image data separated by the processing of the separation step into the predicted image data; A dequantization step of adaptively dequantizing the difference data corresponding to the feature amount of the prediction image data; the decoding, based on the prediction image data and the dequantized difference data. An information processing method comprising: a generation step of generating image data; and an output step of outputting the decoded image data generated in the processing of the generation step. 17. A program for controlling an information processing device that decodes encoded data of original image data into decoded image data corresponding to the original image data, the original image data being further encoded after being encoded. When
Coded image data obtained by coding the predicted image data predicted to be decoded, and the difference between the original image data adaptively quantized with respect to the feature amount of the predicted image data and the predicted image data From encoded data consisting of data,
A separation control step of controlling separation of the coded image data and the quantized difference data, and decoding of the coded image data whose separation is controlled by the processing of the separation control step into the predicted image data. A decoding control step for controlling, a dequantization control step for controlling an adaptive dequantization of the quantized difference data corresponding to a feature amount of the prediction image data, and the prediction image data, A generation control step that controls generation of the decoded image data based on the dequantized difference data; and an output control that controls output of the decoded image data whose generation is controlled by the processing of the generation control step. A recording medium having a computer-readable program recorded thereon, comprising: 18. When the original image data is encoded and then further decoded by a computer that controls an information processing apparatus that decodes encoded data of the original image data into decoded image data corresponding to the original image data. ,
Coded image data obtained by coding the predicted image data predicted to be decoded, and the difference between the original image data adaptively quantized with respect to the feature amount of the predicted image data and the predicted image data From encoded data consisting of data,
A separation control step of controlling separation of the coded image data and the quantized difference data, and decoding of the coded image data whose separation is controlled by the processing of the separation control step into the predicted image data. A decoding control step for controlling, a dequantization control step for controlling an adaptive dequantization of the quantized difference data corresponding to a feature amount of the prediction image data, and the prediction image data, A generation control step that controls generation of the decoded image data based on the dequantized difference data; and an output control that controls output of the decoded image data whose generation is controlled by the processing of the generation control step. A program that causes steps and to be executed.