JPH08307865A - Image processing method - Google Patents

Abstract

PURPOSE: To reduce the redundancy and to improve the compression efficiency by allocating each difference image to a position corresponding to other sampled image so as to utilize the correlation between adjacent picture elements. CONSTITUTION: A picture element value being a mean value of upper left picture elements a-d of sub sampling images B1-B4 is used for an upper left picture element value a' of a mean value image B'1. Furthermore, an upper left image value b' of a difference image B'2 is an upper left picture element value (b) of the image B2 subtracted by the mean value a'. Similarly the upper left picture element c' of the difference image B'3 is a picture element value (c) subtracted by the mean value a', and the picture element value d' of the difference image B'4 is the picture element value (d) of the image B4 by the mean value a'. Similarly a mean value and a difference are calculated as to four picture elements at the same position relatively in the images B1-B4 to obtain an image after conversion. Then the difference images B'1-B'4 are allocated to positions corresponding to the images B1-B4 to calculate the mean value and the difference as to four picture elements at the same position relatively and to obtain an image after the conversion.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method for encoding an image for compression and decoding for decompression,
Particularly, the present invention relates to an image processing method capable of improving compression efficiency in a still image.

[0002]

2. Description of the Related Art JPEG is used as a still image encoding method.
(Joint Photographic coding Experts Group) method is known. The outline of the JPEG method will be described with reference to FIG. FIG. 12 is a schematic block diagram for explaining JPEG encoding / decoding.

First, in JPEG encoding, an original image is divided into blocks of, for example, 8 × 8 pixels, block cosine transform (DCT) is performed on each block, and the transformed coefficients are quantized. (Q), and the quantized coefficient is subjected to variable length coding (VLC) to create coded data. This encoded data is stored in a storage medium or sent to a transmission line.

In the JPEG decoding, the coded data thus encoded is subjected to variable length decoding (VLD), inverse quantization (IQ), inverse discrete cosine transform (IDCT) and block data. To get.

The JPEG system is specifically described in the interface "International Standard Coding System for Color Still Images" by Toshiaki Endo, December 1991, p160-p182.

[0006]

However, in the conventional still image processing method described above, the error in the frequency space increases as the compression rate increases, and the reproduced image deteriorates in decoding for expansion. There was a problem.

Further, it cannot be said that the characteristic that a certain pixel has a high correlation with another adjacent pixel is sufficiently utilized, and the compression efficiency is not improved by utilizing the correlation. was there.

Specifically, as shown in FIG. 13, in general, with respect to an arbitrary one pixel (pixel), there are four neighboring pixels on the upper, lower, left, and right sides, or eight neighboring pixels on the surroundings. It is known that (correlation) is high (similar to each other), and if this correlation can be used, improvement in compression rate can be expected.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an image processing method capable of reducing redundancy by utilizing correlation between adjacent pixels and improving compression efficiency. And

[0010]

The invention according to claim 1 for solving the problems of the above-mentioned conventional example is an image processing for reproducing a still image by encoding a still image and decoding the encoded data. In the method, subsampling is performed by grouping a pixel number of a pixel of the still image by an integer and a quotient and a quotient, and the partial images divided into the groups are used as subsampling images, and all the subsampling images are subjected to relative sampling. The average value of the pixels at a specific pixel position is calculated, a partial image composed of the average value is created as an average value image, and the average value image is assigned to one position of the sub-sampling image. The difference image with each other sub-sampling image is calculated to generate a difference image, and the difference image is compared with the corresponding position of the other sub-sampling image. It is characterized by performing encoding by allocating a.

According to a second aspect of the invention for solving the problems of the conventional example, in the image processing method, the image data encoded by the image processing method according to the first aspect is decoded, and the decoded image data is decoded. Further, the decoded average value image is added to each difference image for each corresponding pixel to form each added image, and each pixel of the average value image is multiplied by an integer for the number of the added image, and all of the above are added. Form a new image in which the difference between the added image and the added image is calculated, and the still image is reproduced by rearranging the pixels of the new image and the added image in the reverse method of subsampling in claim 1. It has a feature.

According to a third aspect of the present invention for solving the problems of the conventional example, in the image processing method, the average value image according to the first aspect is sub-sampled to form a sub-sampling image of the next stage, The average value of the pixels at relative pixel positions is calculated in all the sub-sampling images of the next stage, and a partial image composed of the average values is created as an average value image of the next stage to generate a sub-image of the next stage. Sampling image 1
Assigned to one position to calculate the difference between the next-stage average value image and the other unassigned next-stage sub-sampling image to create the next-stage difference image. Is assigned to the corresponding position of the other sub-sampling image of the next stage, and the average value image of the next stage is obtained by repeating the same process as described above. It is characterized by performing conversion.

According to a fourth aspect of the present invention for solving the problems of the conventional example, in the image processing method, the image data coded by the image processing method according to the third aspect is decoded, and the decoded image data is decoded. The decoded average value image of the final stage is added to each difference image of the final stage for each corresponding pixel to form each added image, and each pixel of the average value image of the final stage is added to the added image. Is multiplied by an integer for several minutes to calculate a difference from all the added images, and a new image is formed, and the new image and the added image are rearranged in pixels by the reverse method of subsampling in claim 1. It is characterized in that the performed images are combined and the combined image is subjected to the same processing as above to reproduce a still image.

[0014]

According to the first and second aspects of the present invention, the original image is subsampled, an average value image and a difference image are formed from the subsampled images, and the images are coded and decoded to obtain the average value. The subsampling image is restored from the image and the difference image, and the pixel processing is performed in the reverse method of subsampling to reproduce the original image, so the correlation between adjacent pixels is used by subsampling. However, the redundancy can be reduced by using the average value image and its difference image.

According to the third and fourth aspects of the present invention, the subsampling and the process of forming the average value image and the difference image in the first aspect are repeatedly encoded in multiple stages, and
The process of decoding according to claim 2 to form a new image and an addition image from the final-stage average value image and the final-stage difference image and combining the images by the reverse method of sub-sampling is repeated in multiple stages to form an image. Since it is an image processing method to reproduce, redundancy is reduced by utilizing the correlation between adjacent pixels by sub-sampling, and by making the average value image at the final stage and the stepwise difference image, it is compressed by the number of repeated processing. The efficiency can be improved.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to the drawings. An image processing method according to an embodiment of the present invention is
The image is divided (subsampling) for each pixel in the vertical direction or the horizontal direction or the vertical and horizontal directions of the original image, and the divided image (subsampling image) is
For the average value image, an average value image composed of average values of pixel values of each sub-sampling image and a difference image obtained by taking a difference from the average value image are created and image conversion for reducing redundancy is performed. Is a sub-sampling, image conversion to reduce redundancy, the above steps are performed in multiple steps, and the final average value image and the stepwise difference image are quantized and variable-length coded. Then, it is possible to improve the compression efficiency by performing coding by utilizing the spatial correlation between adjacent pixels.

First, an image conversion method in the encoding of the image processing method according to the embodiment of the present invention will be described with reference to FIGS.
Use to explain. FIG. 1 is an explanatory diagram showing an example of subsampling in an image conversion method in an image processing method according to an embodiment of the present invention, and FIG. 2 shows reduction of redundancy in the image conversion method of the present embodiment. It is explanatory drawing which shows an example,
FIG. 3 is an explanatory diagram showing an example in which redundancy is reduced in multiple steps in the image conversion method of this embodiment, and FIG. 4 is an explanatory diagram showing the concept of the image conversion method of this embodiment.

The outline of the image conversion method in the encoding of the image processing method according to the embodiment of the present invention is as follows. First, sub-sampling is performed to divide an original image to be encoded into a plurality of images, and Create an average value image from the sub-sampling images divided into a plurality of times, and create a difference image that is the difference between the average value image and the average value image to reduce redundancy and This is a method of further performing subsampling to reduce redundancy, and performing this subsampling and redundancy reduction in multiple stages.

Here, the method of creating the sub-sampling image will be described using three examples. As a first example, when a still image (original image) of one frame is composed of k pixel lines in the vertical direction, the remainder when dividing the line number of the k pixel lines by an integer n The image data is divided into n groups, and pixel data for each line is arranged in each group in the ascending order of the quotient of the division result to create n partial images (sub-sampling images).

As a second example, when the original image is composed of s pixels in the horizontal direction, the pixel number of s pixels is divided into m groups by the remainder when divided by an integer m. Then
Further, the pixel data for each pixel in each group are arranged in the order of increasing quotient of the division result to create m partial images (sub-sampling images).

As a third example, if the original image is composed of s pixels in the horizontal direction and the horizontal pixel line is composed of t pixels in the vertical direction, the pixel number of the s pixels is Is divided by an integer m and the line number of t pixel lines is divided by n to divide into m × n groups, and pixel data for each pixel in each group is divided into The quotients are arranged in ascending order to create m × n partial images (subsampling images).

In the description of this embodiment, as shown in FIG.
An example will be described in which the original image A is divided into 2 × 2 in both the vertical direction and the horizontal direction and four sub-sampling images each having a size of ¼ of the original image are created. That is, for the original image A composed of W pixels in the horizontal direction and H pixels in the vertical direction,
Assuming that each pixel value is p (x, y), the method of creating the sub-sampled images B1, B2, B3, B4 is as follows. B1 = a (i, j) = p (2i, 2j) B2 = b (i, j) = p (2i + 1,2j) B3 = c (i, j) = p (2i, 2j + 1) B4 = D (i, j) = p (2i + 1,2j + 1) where 0≤i <W / 2 and 0≤j <H / 2.

Next, as shown in FIG. 2, the redundancy reduction method in the image conversion method according to the present embodiment, as shown in FIG. 2, averages four pixels in the same position in the sub-sampled images B1, B2, B3, B4. A mean value image B'1 composed of values and difference images B'2, B'3, B'4 between the mean value image B'1 and each sub-sampling image B2, B3, B4 are created. Specifically, the pixel value is calculated using the following formula. B'1 = a '(i, j) = {a (i, j) + b (i, j) + c (i, j) + d (i, j)
} / 4 B'2 = b '(i, j) = b (i, j) -a' (i, j) B'3 = c '(i, j) = c (i, j) -a' (i, j) B'4 = d '(i, j) = d (i, j) -a' (i, j)

For example, the sub-sampling image B1, shown in FIG.
Upper left pixels a (0,0), b (0,0), c (0,0) of B2, B3, B4,
The pixel value obtained by taking the average value from d (0,0) is used as the average value image B'1.
The pixel value a '(0,0) at the upper left of The upper left pixel value b '(0,0) of the difference image B'2 is a value obtained by subtracting the average value a' (0,0) from the upper left pixel value b (0,0) of the sub-sampling image B2. Similarly, the upper left pixel value c ′ (0,0) of the difference image B′3 is
The average value a '(0,0) is subtracted from the upper left pixel value c (0,0) of the sub-sampled image B3, and the upper left pixel value d' (0,0) of the difference image B'4 is , A value obtained by subtracting the average value a '(0,0) from the upper left pixel value d (0,0) of the sub-sampled image B4. Then, similarly, the sub-sampling images B1, B2,
In B3 and B4, the average value and the difference are calculated for the four pixels located at the same relative position, and the converted image is obtained.

Here, the difference images B'2, B'3, B'4 have a large value when the pixel value suddenly changes such as an edge, but in other cases, the value near 0 is set. To take. Therefore,
The amount of information can be reduced by performing quantization and variable length coding on the difference images B'2, B'3, B'4.

On the other hand, the average value image B'1 composed of the average value of 4 pixels may be said to be an image obtained by reducing the original image A to a size of 1/4, and the amount of information is considerably reduced. However, in order to further reduce the amount of information, as a second step, the above-described sub-sampling is performed on the average value image B′1 to further divide the image into four, and the redundancy is reduced as described above. As shown in, the average value image B ″ 1 in the image B′1
And difference images B ″ 2, B ″ 3, B ″ 4 are created. And
Quantization and variable length coding are applied to the difference images B ″ 2, B ″ 3, B ″ 4.

Further, in order to further reduce the amount of information, as a third step, the above-mentioned sub-sampling is performed on the average value image B ″ 1 to divide the image into four, and further the above-mentioned redundancy reduction is performed. As shown in FIG. 3, the average value image B ′ ″ 1 in the image B ″ 1 and the difference image B ′ ″ 2,
B ″ ′ 3, B ′ ″ 4 are created, and the average value image B ′ ″ 1 and the difference images B ″ 2, B ″ 3, B ″ 4 are quantized and variable length code Apply.

In this way, the result of the processing for converting the original image A becomes an image which can be expressed hierarchically as shown in FIG. It should be noted that the number of stages of conversion is arbitrary, and is determined from the balance between the required compression rate and processing speed.

Next, an image inverse conversion method when decoding the encoded data encoded by using the image conversion method of this embodiment will be described with reference to FIG. FIG. 5 is an explanatory diagram showing an example of redundancy restoration in the image inverse conversion method in the decoding of the image processing method according to the present embodiment. Incidentally, FIG.
Then, an example is shown in which the sub-sampling images B1, B2, B3, B4 are restored from the average value image B'1 and the difference images B'2, B'3, B'4.

The outline of the image inverse transform method in the decoding of the image processing method of this embodiment is as follows. First, the redundancy is restored from the average value image and a plurality of difference images to restore the sub-sampling image, and then the restoration is performed. This is a method of reconstructing an original image by performing an operation that is the reverse operation of subsampling from the obtained subsampling image, performing restoration of this redundancy and image synthesis in multiple stages.

First, as shown in FIG. 5, the method of restoring the redundancy is performed by comparing the average value image B'1 and the difference images B'2, B'3, B'4 with four pixels at the same relative position. , The pixel value of the sub-sampling image to be restored is calculated using the following equation. B2 = b (i, j) = b '(i, j) + a' (i, j) B3 = c (i, j) = c '(i, j) + a' (i, j) B4 = d (i, j) = d '(i, j) + a' (i, j) B1 = a (i, j) = 4 * a '(i, j) -b (i, j) -c ( i, j) −
d (i, j)

For example, as shown in FIG. 5, the sub-sampling image B2 has a pixel value b '(i, j) of the difference image B'2 and a pixel value a' (i, j) of the average value image B'1. And the sub-sampling image B3 is similar to the difference image B'3.
Pixel value c ′ (i, j) of the average value image B′1 and the pixel value a ′ (i, j) of the average value image B′1 are added, and the subsampling image B4
Is composed of a pixel value obtained by adding the pixel value d ′ (i, j) of the difference image B′4 and the pixel value a ′ (i, j) of the average value image B′1. That is, the sub-sampling images B2, B3, B4 are addition images obtained by adding the difference images B'2, B'3, B'4 to the average value image B'1.

Finally, the sub-sampled image B
1 is the pixel value b (i, j) of the sub-sampling image B2 and the pixel of the sub-sampling image B3 restored from the pixel value obtained by multiplying the pixel value a '(i, j) of the average value image B'1 by 4. Value c (i,
j) and the pixel value d (i, j) of the sub-sampled image B4. That is, the subsampling image B1 is the average value image B'1 and the subsampling image B1.
A new image is formed from 2, B3, B4.

Next, the image synthesizing method for synthesizing the original image from the restored sub-sampled image is an operation reverse to the sub-sampling in the above-mentioned encoding, so that the pixels of each sub-sampled image are in the original order. In this embodiment, an image having a size four times larger is created. Specifically, the restored sub-sampled images B1, B2, B3, B4 are combined to generate the original image A.

The details of the inverse transform method will be described here in only one stage of inverse transform. However, in the case of performing the transform in multiple stages in the image transform of the encoding unit, for example, an example shown in FIG. 4 will be described. , The last-obtained average value image B ′ ″ 1 and the difference images B ′ ″ 2, B ′ ″ 3, B ′ ″ 4 are inversely transformed to obtain the average value image B ″ 1 at the previous stage. Generate the average value image B ″ 1 and the difference images B ″ 2, B ″ 3, B ″ 4 of the previous stage to generate the average value image B′1 of the previous stage, and repeat this. The original image A is generated.

Next, an encoding apparatus for implementing the image processing method of this embodiment will be described with reference to FIG. FIG. 6 is a block diagram of the configuration of an encoding device that realizes the image processing method of this embodiment. As shown in FIG. 6, the encoding apparatus according to the present embodiment includes an image converter 11, a quantizer 12, a variable length encoder 13, and a multiplexer (MPX) 1.
And 4.

Next, each section of the coding apparatus of this embodiment will be specifically described. The quantizer 12 quantizes using a quantization table, and the variable length encoder 13
Is to be encoded by using an encoding table, and is realized by using a generally known technique. The quantizer 12 and the variable length encoder 13 are
Four input / output terminals are provided, and a maximum of four data inputs / processes / outputs can be performed in parallel.
The MPX 14 temporarily stores and combines the encoded data output from the four output terminals of the variable length encoder 13,
An error correction code or the like is added, and the data is packetized and output.

The image converter 11, which is a characteristic part of the present embodiment, performs subsampling by utilizing the correlation between pixels described in FIGS. 1 to 4, and performs conversion for reducing redundancy in multiple stages. It is something that is repeated. Further, the circuit configuration inside the image converter 11 will be described with reference to FIG. FIG.
It is a schematic circuit block diagram of the image converter 11 of a present Example.

As shown in FIG. 7, the image converter 11 includes an image divider 21, a 4-input adder 22, a 1 / 4-fold amplifier 23, 2-input difference units 24, 25 and 26, and switches ( SW) 27.

The image divider 21 includes a memory (large memory) for temporarily storing the image (original image) data inputted therein for one screen, and four sub-images obtained by sub-sampling the original image 1. It has a memory (four small memories) for storing each sampled image. Then, the image divider 21 creates a sub-sampling image from the original image stored in the large memory by using the above-described sub-sampling method, stores the sub-sampling image in each small memory, and outputs the images of the sub-sampling images all at once from the output terminals. The data is output. When the sub-sampling of the original image in the large memory is completed, the large memory is reset, and when all the pixel data (pixel value data) of the image in the small memory is output from the output terminal, the small memory is reset. It is supposed to be done.

Specifically, the original image A shown in FIG. 1 is stored in the large memory.
Are stored, subsampled and subsampled images B1, B2, B3, B4 are stored in a small memory. Furthermore,
Pixel data a of sub-sampled images B1, B2, B3, B4
(0,0), b (0,0), c (0,0), d (0,0) are first output terminals (1) (2)
It will be output from (3) and (4).

The 4-input adder 22 has output terminals (1) (2) (3)
The pixel data of (4) is added, the 1 / 4-fold amplifier 23 calculates the added pixel data by 1/4, and the SW 27 is
Only when outputting the minimum unit of the average value image, the connection is made so as to output to the output 1, but otherwise, the connection is made so as to be fed back and inputted to the image divider 21.

The 2-input differentiator 24 is a 1/4 amplifier 23.
The negative value of the output (pixel data of the average value image B'1) and the value of the output (pixel data of the sub-sampling image B2) from the output terminal (2) of the image divider 21 are added to obtain both data. The difference (pixel data of the difference image B′2) is output.

The two-input difference unit 25 is similar to the two-input difference unit 24 and outputs the difference between the pixel data of the average value image B′1 and the pixel data of the sub-sampling image B3 as the pixel data of the difference image B′3. The two-input differentiator 26
Is the same as the two-input differentiator 24, 25, and the average value image B'1
The difference between the pixel data of the sub-sampling image B4 and the pixel data of the sub-sampling image B4 is output as the pixel data of the difference image B'4.

Next, the processing of the image coding apparatus of this embodiment will be specifically described. Here, the case where the average value image B ′ ″ 1 in FIG. 4 is the minimum average value image will be described.
First, the image data of the original image A is input to the image divider 21 and stored in a large memory, and subsampling is performed by the operation of the image divider 21 to obtain the subsampled image B1 ,.
B2, B3, and B4 are stored in the small memory, the large memory is reset, and the contents are cleared.

Then, from the output terminals (1) (2) (3) (4), the image B
Pixel data a (0,0), b (0,1) of 1, B2, B3, B4
0), c (0,0), d (0,0) are simultaneously output, four pixel data are added by the 4-input adder 22, and the 1/4 times amplifier 2
It is amplified by 1/4 at 3. That is, 1/4 times amplifier 2
The output from 3 becomes the average value of the four pixel data, and when all the pixel data in the sub-sampling image are subjected to the above processing, the average value image B′1 is formed.

Then, the difference between the pixel data of the average value image B'1 and the pixel data of the sub-sampling image B2 is calculated by the 2-input difference device 24, and the pixel data of the average value image B'1 and the pixel data of the sub-sampling image B3 are calculated. The difference between the pixel data of the average value image B′1 and the pixel data of the sub-sampling image B4 is calculated by the 2-input difference device 26, and the difference image B is output to outputs 1 to 4. '1, B'2, B'3 are output and input to the quantizer 12. However,
Since SW27 is not connected to the output 1 side at this time,
The average value image B'1 is not output to the output 1 and the image divider 2
Input to 1. The sub-sampling images B1, B2, B
When all the pixel data of 3, B4 are read, the small memory is reset and the contents are cleared.

Next, the image data of the average value image B'1 input to the image divider 21 is stored in a large memory, further subsampled by the image divider 21, and divided into subsampled images. The pixel data of the pixel at the corresponding position in each sub-sampling image is output from the output terminals (1) (2) (3) (4). The average value image B ″ 1 and the difference images B ″ 2, B ″ 3, B ′ are operated by the operations of the 4-input adder 22, the 1 / 4-fold amplifier 23, and the 2-input difference devices 24, 25, and 26. '4 is obtained, the average value image B ″ 1 is fed back to the image divider 21, and the difference images B ″ 2, B ″ 3, B ″ 4 are inputted to the quantizer 12. It

When the above process is repeated to reach the stage of obtaining the average value image B '''1 which is the minimum average value image, S
W27 switches to the output 1 side and outputs the obtained average value image B ′ ″ 1 and difference images B ′ ″ 2, B ′ ″ 3, B ″ ′ 4 to the quantizer 12. is there. In this way, the image conversion as shown in FIG. 4 is performed. Then, the image-converted data is quantized and variable-length coded to obtain MPX.
14 is output to the MPX 14, and the MPX 14 temporarily stores the same array of encoded data as in the state shown in FIG. 4, error-correction-encodes the content, packetizes it, and outputs it as encoded data to a transmission line or the like. .

The order of encoding in the above processing is shown in FIG.
Use to explain. FIG. 8 is an explanatory diagram showing the minimum unit of encoding. As shown in FIG. 8, the difference images B′2, B′3,
B′4 is simultaneously encoded in 8 × 8 pixel block units, then the difference images B ″ 2, B ″ 3, B ″ 4 are simultaneously encoded in 8 × 8 pixel block units, and then , The difference image B ′ ″ 2,
B ″ ″ 3 and B ″ ″ 4 are simultaneously encoded in 8 × 8 pixel block units, and finally, the average value image B ′ ″ 1 is encoded in 8 × 8 pixel block units.

The outputs 1 to 4 in FIG. 7 are not directly input to the quantizer 12 but are temporarily stored in the memory for one screen in the image converter 11 in the state shown in FIG. As shown in FIG. 9, pixel data corresponding to the position and size of the average value image and the difference image are selected, and 8 ×
You may make it encode by forming an 8 pixel block.

In this case, quantization and variable length coding are performed in units of 8 × 8 pixel blocks, and M blocks are coded in the order of coding.
8 if packetized by PX14 and output
Since the × 8 pixel block contains only one important pixel data of the average value image, when an error occurs in the transmission after encoding by the MPX 14, the influence of the error can be minimized. There is an effect that can be done.

Next, a decoding device for realizing the image processing method of this embodiment will be described with reference to FIG. Figure 10
FIG. 3 is a configuration block diagram of a decoding device that realizes the image processing method of the present embodiment. The decoding apparatus of this embodiment is shown in FIG.
As shown in, the demultiplexer (DM)
PX) 31, variable length decoder 32, and inverse quantizer 33
And an image inverse converter 34.

Next, each section of the decoding apparatus of this embodiment will be concretely described. The variable length decoder 32 performs decoding using the same encoding table as that of the variable length encoder 13 of the encoding device, and the inverse quantizer 33 is equivalent to the quantizer 12 of the encoding device. Inverse quantization is performed using the same quantization table, and all of them are implemented using commonly known techniques.

The DMPX 31 extracts coded data from the packetized coded data, performs error correction and the like, and outputs it to the variable length decoder 32. However, the DMPX 31 of this embodiment synthesizes the received encoded data and arranges the encoded data in the MPX 14 (see FIG. 4).
The state of arrangement of encoded data is equal to the state of.

Next, the image inverse converter 34, which is a characteristic part of this embodiment, will be described. The image inverse converter 34 is shown in FIG.
The redundancy is restored from the average value image and the difference image described in 1. to restore the sub-sampling image, and the sub-sampling images are combined to reproduce the original image.

The image inverse converter 34 of this embodiment is shown in FIG.
A specific description will be given using 1. FIG. 11 is a schematic circuit configuration diagram of the image inverse converter 34 of the present embodiment. The image inverse converter 34 of the present embodiment, as shown in FIG.
W) 41, 2-input adders 42, 43, 44, quadruple amplifier 45, 4-input difference unit 46, image synthesizer 47,
It is composed of a switch (SW) 48.

It should be noted that the input 1 of the image inverse converter 34 of this embodiment is
4 to the variable length decoder 32 and the inverse quantizer 33 so that the encoded data encoded in the order described in FIG. The converted image data is input, and specifically, first, the average image B ′ ″ 1 and the difference images B ″ ′ 2 and B ″ ′ 3 of the minimum unit are first input. , B '''4 is input 1,
It is designed to be input to 2, 3, and 4.

SW41 has a switch connected to the input 1 side when an average value image of the minimum unit is input to the input 1.
Others are connected to the image synthesizer 47 side. The SW 48 is connected to the output side when an image for one screen is reproduced (when the image reverse conversion is completed), but is connected to the SW 41 side for the others.

The 2-input adder 42 adds the pixel data of the average value image B ′ ″ 1 and the pixel data of the difference image B ′ ″ 2, and the 2-input adder 43 outputs the average value image B ′. The pixel data of "" 1 and the pixel data of the difference image B '"3 are added, and the 2-input adder 44 uses the pixel data of the average value image B'" 1 and the difference image B "'. The pixel data of 4 is added.

The output from the 2-input adder 42 becomes the pixel data of the sub-sampling image based on the difference image B ′ ″ 2, and the output from the 2-input adder 43 becomes the difference image B ″ ′ 3.
It becomes the pixel data of the sub-sampling image based on
The output from the 2-input adder 44 becomes the pixel data of the sub-sampling image based on the difference image B ′ ″ 4.

The quadruple amplifier 45 amplifies the pixel data of the average value image B ′ ″ 1 four times, and the four-input difference device 46 outputs the four-fold amplified values to the two-input adders 42, 43 and 44. The difference is obtained by subtracting the value of. This difference value becomes the pixel data of the sub-sampling image based on the average value image B ′ ″ 1.

The image synthesizer 47 has terminals for four inputs and one output, and each has a small memory for storing pixel data from the four inputs, and in the above example, the average value image B ''' 1 based sub-sampling image, difference image B '''2 based sub-sampling image, difference image B''' 3 based sub-sampling image, difference image B ''' A sub-sampling image based on 4 is stored.

Then, the image synthesizer 47 synthesizes the sub-sampled images by the reverse method of the sub-sampling and temporarily stores them in the large memory. Further, the contents of the large memory are sequentially output to SW41 via SW48, and SW41
Is connected to the one other than the input 1, so that its contents are input to the 2-input adders 42, 43, 44 and the quadruple amplifier 45.

The DMPX 31 selects and inputs the difference images B ″ 2, B ″ 3, B ″ 4 to the inputs 2, 3 and 4, and the 2-input adders 42, 43 and 44. And 4x amplifier 4
The same processing as above is performed by the 5 and 4-input subtractor 46,
A composite image of the sub-sampling images is used as an average value image, a sub-sampling image based on the average value image, a sub-sampling image based on the difference image B ″ 2, and a difference image B ″ 3. The sub-sampling image and the sub-sampling image based on the difference image B ″ 4 are obtained.

Further, the DMPX 31 selects and inputs the difference images B'2, B'3, B'4 to the inputs 2, 3 and 4, and the 2-input adders 42, 43, 44 and 4 are used. Double amplifier 45
And the four-input subtractor 46 perform the same processing as described above to make the composite image of the sub-sampling images the average value image,
A sub-sampling image based on the average value image, a sub-sampling image based on the difference image B′2, a sub-sampling image based on the difference image B′3, and a difference image B′4. And a sub-sampled image is obtained. As described above, the image is converted from the average value image and the difference image by performing the image inverse conversion using a method reverse to the image conversion.

According to the image processing method of the present embodiment, a small sub-sampling image obtained by sub-sampling an original image is divided to calculate an average value and a difference value of pixels having a relative positional relationship, and one is calculated as an average value image. , The other is assigned to the difference image, image conversion is further performed, the average value image is subsampled, and the average value image and the difference image are processed by the same image conversion process as described above. Since the average value image at the final stage) and other stepwise difference images are encoded, and the image is reproduced by performing decoding and image inverse conversion, the spatial correlation between pixels is used to generate the difference image. By creating a large number of parts, the amount of information is reduced, redundancy is reduced, and encoding and decoding are performed with an image with a small amount of information, which has the effect of improving the compression efficiency of the image.

Furthermore, when the image processing method of this embodiment is used, the amount of encoded data is reduced, so that the transmission time is shortened when transmitting encoded data, and the transmission error can be reduced. is there.

[0069]

According to the first and second aspects of the present invention, the original image is sub-sampled, the average value image and the difference image are formed from the sub-sampled images, and the images are encoded and decoded. The sub-sampling image is reconstructed from the average value image and the difference image, and the pixel processing is performed in the reverse method of sub-sampling to reproduce the original image. By using the average value image and the difference image thereof, it is possible to reduce the redundancy.

According to the third and fourth aspects of the invention, the subsampling and the process of forming the average value image and the difference image in the first aspect are repeatedly encoded in multiple stages, and
The process of decoding according to claim 2 to form a new image and an addition image from the final-stage average value image and the final-stage difference image and combining the images by the reverse method of sub-sampling is repeated in multiple stages to form an image. Since it is an image processing method to reproduce, redundancy is reduced by utilizing the correlation between adjacent pixels by sub-sampling, and by making the average value image at the final stage and the stepwise difference image, it is compressed by the number of repeated processing. There is an effect that efficiency can be improved.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing an example of subsampling in an image conversion method in encoding of an image processing method according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an example of reduction of redundancy in the image conversion method of the present embodiment.

FIG. 3 is an explanatory diagram showing an example in which redundancy is reduced in multiple steps in the image conversion method of the present embodiment.

FIG. 4 is an explanatory diagram showing the concept of the image conversion method of the present embodiment.

FIG. 5 is an explanatory diagram showing an example of redundancy restoration in the image inverse conversion method in the decoding of the image processing method according to the embodiment.

FIG. 6 is a configuration block diagram of an encoding device that realizes the image processing method of the present embodiment.

FIG. 7 is a schematic circuit configuration diagram of an image converter 11 of the present embodiment.

FIG. 8 is an explanatory diagram showing a minimum unit of encoding.

FIG. 9 is an explanatory diagram showing a minimum unit of encoding.

FIG. 10 is a configuration block diagram of a decoding device that realizes the image processing method of the present embodiment.

FIG. 11 is a schematic circuit configuration diagram of an image inverse converter 34 of the present embodiment.

FIG. 12 is a schematic block diagram for explaining JPEG encoding / decoding.

FIG. 13 is an explanatory diagram illustrating the correlation of pixels.

[Explanation of symbols]

11 ... Image converter, 12 ... Quantizer, 13 ... Variable length encoder, 14 ... Multiplexer, 21 ... Image divider, 22 ... 4-input adder, 23 ... 1/4 times amplifier,
24, 25, 26 ... 2 input difference device, 27 ... Switch, 31 ... Demultiplexer, 32 ... Variable length decoder, 33 ... Inverse quantizer, 34 ... Image inverse converter, 4
1 ... switch, 42, 43, 44 ... 2-input adder,
45 ... 4 times amplifier, 46 ... 4-input difference device, 47 ... Image synthesizer, 48 ... Switch

Claims (4)

[Claims] 1. An image processing method for encoding a still image, decoding the encoded data and reproducing the still image, wherein the pixel numbers of the pixels of the still image are grouped by a remainder and a quotient divided by an integer. Sub-sampling is performed, the partial images divided into the groups are used as sub-sampling images, the average value of pixels at relative pixel positions in all the sub-sampling images is calculated, and the partial image composed of the average values is calculated. As an average value image and assign it to one position of the sub-sampling image, calculate a difference between the average value image and another sub-sampling image that is not assigned to create a difference image, and An image processing method, wherein a difference image is assigned to a position corresponding to the other sub-sampling image and encoded. 2. The image data encoded by the image processing method according to claim 1, is decoded, and the decoded average image is added to each of the decoded difference images for each corresponding pixel. Each addition image is formed, each pixel of the average value image is multiplied by an integer by the number of the addition image to form a new image in which the difference from all the addition images is calculated, and the new image and the new image are formed. An image processing method characterized in that a pixel is rearranged in the added image by a method reverse to the sub-sampling in claim 1 to reproduce a still image. 3. An average value image according to claim 1 is sub-sampled to form a sub-sampling image of a next stage, and an average value of pixels at relative pixel positions in all the sub-sampling images of the next stage is calculated. A partial image composed of the average values is calculated as a next-stage average value image, and is assigned to one position of the next-stage sub-sampling image, and is not assigned to the next-stage average value image. To calculate a difference with the next-stage sub-sampling image to create a next-stage difference image, and assign each of the next-stage difference images to a corresponding position of the other next-stage sub-sampling image, and An image processing method, wherein the average value image at the final stage obtained by repeating the same process as the average value image at the next stage and a stepwise difference image are encoded. 4. The image data encoded by the image processing method according to claim 3, is decoded, and each of the decoded differential image of the final stage is associated with the average image of the decoded final stage. Addition is performed for each pixel to form each addition image, and each pixel of the final-stage average value image is multiplied by an integer for the number of the addition image to calculate a new image in which the difference from all the addition images is calculated. Forming the new image and the added image
An image processing method characterized in that pixels are rearranged by the reverse method of sub-sampling to synthesize images, and the synthesized image is repeated to reproduce a still image.