JPH0937262A - Image processing unit and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain coding for image data in which an error of picture elements is controlled at a high compression efficiency by converting a picture element or a block in an image so as to use a comparatively simple circuit configuration or processing means. SOLUTION: At first, a reference Huffman table is set (S0). Image data are read and block data are extracted from the image data by a maximum block size (16×16 picture elements) (S1, S2). Then a block division method and a coding method are decided and they are stored as reference information (S3, S4). A statistic quantity is calculated for frequency of incidence of an outputted difference (prediction error) according to the reference information (S5) Block data extracted by the S2 are rewritten into data of a virtual block (S6). An end of the image is discriminated (S7). Then an optimum Huffman table is generated based on the difference statistic value obtained in the S5 (S8). Then an end of loop is discriminated (S9), the entire image is coded based on the reference information (S10).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus and method for encoding an image.

[0002]

2. Description of the Related Art An image encoding method is an information-preserving encoding method that completely restores an original image upon decoding, and an information-non-preserving encoding method that causes an error between the original image and the original image. There are two main methods.

DPCM coding is widely known as an information storage type coding method. According to general DPCM coding, coded pixels around a pixel to be coded, that is, a left pixel (this is referred to as A), an upper pixel (similarly B), and a left diagonal pixel (similarly C). ) Is used to calculate the predicted value of the pixel to be encoded using, for example, the formula (A + BC), and the difference between this and the pixel value to be encoded is encoded.

Although the information non-conservation type encoding method cannot completely restore the original image, it can obtain a high compression rate, so that it handles a large amount of image data and some errors. If is not a problem, an information non-conservation type encoding method is often used. The JPEG baseline method is one of information-non-preserving encoding methods.

[0005]

In the conventional image coding, the amount of information is controlled according to the amount of coded data. Therefore, the actual pixel value of the image to be coded and this Encodes the pixel value error within a certain value with high compression efficiency for an image that allows an error with the pixel value of the decoded image to some extent (to limit the error range). There was a problem that I could not.

In order to solve the above problems, the present invention is
It is an object of the present invention to encode image data with high compression efficiency and controlling an error in pixel value.

Another object of the present invention is to encode multi-valued data with high compression efficiency and by controlling the error of pixel values.

[0008]

In order to solve the above-mentioned problems, an image coding apparatus according to claim 1 of the present invention is an image processing apparatus which performs predictive coding of an image. Predicting means for predicting, an image converting means for converting the pixel value of the image within a range of ± e based on the result of the prediction by the predicting means, an image converted by the image converting means, and the predicting means. Coding means for coding the parameters relating to the image based on the result of the prediction.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION

(First Embodiment) Next, the present invention will be briefly described.

FIG. 1 shows a processing flow of the first embodiment according to the present invention. In the figure, S0 is a step of setting a reference Huffman table described later, S1 is a step of reading image data, S2 is a step of extracting block data from the image data with a maximum block size (here, 16 × 16 pixels), and S3 is a block. A step of deciding a division method and an encoding method, a step S4 of storing the block division method and the encoding method decided in S3 as reference information, and a step S5 of outputting a difference value (prediction error) described later according to the reference information. S6 is a step of taking statistics of appearance frequency, S6 is a step of rewriting the block data extracted in S2 into data of a fictitious block described later, S7 is a step of determining the end of the image, and S8 is based on the difference value statistics obtained in S5. The step of creating the optimum Huffman table, S9 is a step for judging the end of the loop. Step, S9
Is a step of performing encoding processing of the entire image based on the reference information.

The present embodiment will be briefly described. The encoding according to the present embodiment is not lossy encoding, but is an encoding method that allows an error between a pixel value indicated by an actual image and a pixel value of an image to be encoded within a maximum range of ± e. This is called near lossless coding.

Here, the maximum size of the block to be encoded is 16 × 16 pixels, and 16 × 16 pixels, 8 × 8 pixels, 4 ×
Encoding processing is performed with any of the four pixels. As a coding method for each block, six types of predictive coding methods (# 0 to # 5) having different prediction methods and two types of plane coding methods (#
6 and # 7), and the run length coding method (# 8) that can be used only when the inside of the block is binary (see FIG. 2).

First of all, these 9 are used for coding the block.
The types of encoding methods will be briefly described. FIG. 3 shows the structure of code data by each coding method.

# 0 to # 5 are predictive coding methods, which are A + BC, B, A, B + (AC) using the peripheral pixels A, B, C shown in FIG. 2A, respectively. ) / 2, A + (B
-C) / 2, (A + B + C) / 3 is used as a prediction value, and the difference between the pixel of interest and the prediction value is Huffman-encoded. The Huffman table is prepared for each block size, and is encoded using different Huffman tables according to the block size.

# 6 and # 7 are plane coding methods.
In # 6, the pixel value in the upper left corner of the block is c0,
Plane P (x, y) = ax + approximating the pixel value of the block
By + c0 is obtained, and the values of a, b, and c0 and the difference between this approximate plane and the pixel value in the block are Huffman-encoded.

At this time, when a and b are within a certain value, c0 is Huffman-coded for the difference c0 'from the previously appearing c0, and when it is more than a certain value, the value of c0 itself is encoded. . Further, a flag is provided after a, b and c0 so that the difference data is not added when all the difference values in the block are 0.

# 7 has the same pixel value in the block,
That is, it is used when the plane is a = b = 0, and
Huffman code c0 '. In Figure 4, c in # 6 and # 7
An example of a Huffman table used when encoding 0'is shown. In the figure, "S" represents a sign bit, and 0 or 1 is entered depending on the sign of the difference value of c0. Also,
If c0 is greater than or equal to 14 and less than or equal to −14, the difference value is not used and E
The value of c0 shall be added after the SC code.

# 8 is a run-length coding method and is used only when there are only two pixel values in a block. For example, if the pixel with the higher luminance is white and the pixel with the lower luminance is black, the pixels in the block are scanned in the order shown in FIG. 5, converted into run lengths, and encoded.
A Huffman table for a white (high brightness value) run and a black (low brightness value) run that are determined in advance is used for encoding, and the last run in the block is replaced with an EOB code for encoding.

Next, the flow of processing in this embodiment will be described in detail with reference to FIG.

First, in step S0, a reference Huffman table used in block coding is set. This is because the Huffman table created based on the predictive coding obtained in advance for a plurality of images and the statistical information of the prediction error that occurs in other images is read, and the Huffman table for predictive coding of 8 × 8 pixels of 16 × 16 pixels is read. A prediction coding Huffman table, a 4 × 4 pixel prediction coding Huffman table, and a plane coding Huffman table are set.

Further, four counters used when these Huffman tables are newly created in S8, that is, a 16 × 16 pixel predictive coding counter, an 8 × 8 pixel predictive coding counter, and a 4 × 4 pixel predictive coding counter The values of the predictive coding counter and the plane coding counter are all initialized to 0.

Further, a Huffman table for Huffman-encoding the information indicating the encoding method (# 0 to # 8) is prepared for each block size, and is used when these Huffman tables are newly created in S8. An encoding number counter is prepared, and the value of this counter is also set to 0.

In step S1, the image data to be encoded is read into the memory for encoding processing.

In step S2, block data (maximum block size 16 × 16 pixels) is sequentially extracted from the image to be encoded. However, one line above and to the left of the extracted block is also extracted as a reference area for predictive coding.

In step S3, with respect to the maximum size block (16 × 16 pixels) extracted in step S1, the block is further divided into blocks of a size suitable for encoding (encoding target block), and The encoding method (# 0 to # 8) suitable for the encoding target block is determined, and the pixel value in the maximum block is rewritten within the range of an error ± e from the actual pixel value in accordance with this. This makes it possible to reduce the difference value (prediction error) for encoding. Further, by rewriting within a range of ± e so that the difference value has only several kinds of values, the information amount of the difference value can be reduced. A detailed description of the process of step 3 will be given below.

The routine of step S3 is a routine having a recursive structure, and is a maximum block or a square block obtained by dividing this block several times (this block is called a block to be processed, and its size varies depending on the flow of processing). .)
Is divided into sub-blocks (4 divisions) and the code amount without division is compared to select whether to divide the block (see FIG. 6). The reference information including the information information indicating the information and the code amount when the block selected based on the reference information is encoded are output.

FIG. 7 shows in detail the flow of the iterative processing used when selecting the coding method and the division method for the largest block (block to be processed) in step S3.

In FIG. 7, S30 is a step for checking the size of the block to be processed. If the block to be processed is the minimum block size (4 × 4 pixels), the process proceeds to S33, and if not, the process proceeds to S20 and S23.

In step S20, the block to be processed is copied to generate block A. The block to be processed is set as block A and is copied to the memory area for this iterative process. S21 is a step of selecting an optimum coding method that minimizes the code amount LA when the block A is coded. Details of this selection method will be described later.

S22 is a step of outputting the generated code amount LA when the encoding method selected in S21 is used.

Further, S23 is a step of copying the block to be processed to generate the block B, as in S20.
S24 is a step of dividing the block B into four to generate sub-blocks, and S25 is a step of selecting an optimum coding method (optimal coding method) for each sub-block as in S21.

S26 is a step of outputting the total code amount LB, which is the sum of the code amounts when each of the four sub-blocks is encoded using the encoding method selected in S25.

At S27, the LA output at S22 and S29 are output.
Compare the LB output in step S28. If LA <LB, go to S28.
If LA ≧ LB, the step proceeds to S29.

In S28, the size of the block to be processed of block A is determined as the size of the block to be coded, and the iterative process for determining the block division method for the range of this block to be processed is completed.
In S28 of this iterative process, the block A, that is, the target block to be encoded, is the pixel value (range of error ± e) of an imaginary block, which will be described later, generated when the encoding method is selected in S21.
Is rewritten as Further, the process proceeds to S31, and the block division information and the coding method and code amount LA selected in S21 are generated as reference information.

In S29, the block B composed of four sub-block blocks is copied to the range of the block to be processed used at the beginning of this iterative process.
Each of the four sub-blocks is replaced with the block to be processed, and the iterative process for determining the block division method is repeated from the beginning.

In S32, the optimum coding method for the block to be processed having the minimum block size is selected. In S33, the code amount L at that time is output. In S34, the block to be coded is rewritten with the pixel value of an imaginary block, which will be described later, generated in S32. In S35, the optimum coding method and the code amount L corresponding to the block are generated.

This is the end of the description of the flow of the iterative process used when selecting the coding method of the largest block (or the block to be processed) and the block division method in step S3.

Next, the selection of the optimum coding method in steps S21, S25 and S32 will be described in detail.

FIG. 4 is a diagram showing the flow of selecting the optimum coding method. Here, a process of estimating the code amount of all selectable encoding methods and selecting an optimal encoding method is performed.

Before this code amount estimation processing, the block data is rewritten in a form suitable for each method. In FIG. 8, S40 is the code amount L generated by the coding method # 0.
The step of initializing n (0), S41 is the predicted value A + B-
The step of obtaining C, S42 is the step of rewriting the pixel value in the block within the range of error ± e, S43 is the step of adding the Huffman code length for the difference value to Ln (0) by one pixel, and S44 is the end of the block. If the code amount when the block is encoded can be generated, S
Go to 61.

If this processing is set to mode0, mode1 to mode
5 is a process corresponding to each of the coding methods # 1 to # 5, and performs the same process as described above.

S45 is an initialization step relating to the coding methods # 6 and # 7, which is a flag (F) indicating whether or not the pixel values in the approximate plane and the block completely match.
LAG) is prepared and this is initially set to 1 (complete match).
Also, a parameter Ln (6) indicating the code length for # 6
Is set to 0 and the parameter Ln indicating the code length for # 7 is set.
(7) is set to a value that does not become the minimum value among Ln (0) to Ln (8). S46 is a step of determining whether or not the pixel value in the block can be made the same pixel value if the pixel value in the block is changed within an error range of ± e. S47 is the block approximation plane P (# 6) of # 6 described above. x, y) = a
The step of obtaining x + by + c0, S48 is a pixel-by-pixel step of rewriting the pixel value within the error range of ± e to generate the pixel value for forming the fictitious block, and S49 is the generated pixel value and the approximate plane P (x, This is a step of adding the Huffman code length corresponding to the difference value from y) to Ln (6).
S50 is a step of determining whether or not the difference value (prediction error) is 0. If even one pixel value different from the approximation plane exists in the block, FLAG is set to 0 in S51. To be done. The end of the block is determined in S52.

S53 is a step for determining the value of FLAG. If FLAG = 0, the code amount of the difference value for each pixel generated in S49 and the parameter a, which indicates the approximation plane P (x, y), The code amounts of b and c0 (c0 ') are added to obtain Ln (6) (S54), and the process proceeds to S61.

If FLAG = 1 (the approximation plane and the block pixel values are completely the same), only the information of the approximation plane needs to be encoded, so the code amount of the parameters a, b, c0 (c0 ') is set. Replace with Ln (6) (S55), and proceed to S61.

When the value of Ln (6) is obtained, L
n (7) is set to a value that does not become the minimum value among Ln (0) to Ln (8).

In S56, the code amount of c0 (c0 ') is set to the value of Ln (7), and Ln (6) is set to a value that does not become the minimum value among Ln (0 to Ln (8). Then, in S64, the block data is rewritten within the range of ± e (a fictitious block is generated) so that the pixel values in the block are all c0, and the process proceeds to S61.

S57 is a step of determining whether or not the inside of the block is represented by two pixel values. If it is represented by two pixel values, the process proceeds to S58, and if different, Ln (8 ) Is set to a value that does not become the minimum value among Ln (0) to Ln (8), and the process proceeds to S61.

Step S58 is a step of converting the pixel value into a run length. In S59, the generated code amount Ln (8) when this run length is encoded is obtained, and the flow proceeds to S61.

S61 is a step of comparing the values of Ln (0) to Ln (8) and selecting the minimum value and the corresponding optimum coding method. S62 is for each mode (mode0-mode
8) is a step of saving a virtual block corresponding to the encoding method selected in S61 among the virtual blocks generated by rewriting the pixel values in the range of ± e, and S63
Outputs the optimum coding method and code amount.

The process up to step S61 is to obtain the code amount of each method, but these processes are performed after copying the block (or sub-block) to be processed into a different memory area. Therefore, the rewriting of block data when selecting the optimum coding method does not affect the original block to be processed.

The steps S40 to S44 will be described in detail. In steps S40 to S44, the generated code amount L by # 0
This is to obtain n (0).

S41 to S44 are loops for rasterizing the pixels in the block in order from the upper left, and the processed pixels (S
The pixel rewritten in 42) is replaced with the original pixel value as a reference pixel of the target pixel thereafter. Reference pixel (A,
B, C) is outside the area of the block of 16 × 16 pixels extracted in S2, the pixel value of the block data rewritten in S6 by the loop processing (S2 to S7) of the preceding largest block is Will be used.

First, in step S40, 0 is set to Ln (0). In step S41, a predicted value (P) = A + BC is generated from the pixel value X of interest and peripheral pixel values A, B, and C, and a difference value (D) = X-P is obtained. In step S42, the modified difference value (denoted as D ') is obtained from the equation (1), and the pixel of interest is rewritten to D' + P. (1)
In the formula, ↓ x ↓ represents rounding down after the decimal point.

D ′ = ↓ (D + α) / (2e + 1) ↓ × (2e + 1) ... (1) α = (D / | D |) × e ... (2) In FIG. 11, e = 3. This is an example of the modified difference value D ′ when the existing difference value D becomes 4. As shown in FIG. 11, compared to the number of difference values D (-10 to 10), the difference value D '(-
Since the number of 7, 0, 7) is smaller, the information indicating the difference value is greatly reduced. Moreover, the possibility that the difference value completely matches the predicted value is increased.

In this example, the pixel value to be corrected (assumed to be the corrected pixel value X ') = X + 3. In step S43, the difference value between the rewritten corrected pixel value X'and the predicted value P, that is, the corrected difference value D'is Huffman-encoded by the Huffman table for the size of the block (or sub-block) to be processed, and generated. The code length is added to Ln (0). In step S44, it is determined whether or not the processing is completed up to the end of the block. If it is completed, the processing is moved to step S61,
If the process is not finished, the process from step S41 is repeated focusing on the next pixel. As a result, the generated code amount Ln (0) when the # 0 coding method is used is obtained. Also,#
No. 1 to No. 5 are different only in the prediction method, and the method of obtaining the generated code amount Ln (1 to 5) is the same as that of No. 0, so detailed description will be omitted.

Next, steps S45 to S56 and S64 will be described in detail. Step S45 is a step for initializing the variables for # 6 and # 7. The code amount Ln (6) generated by # 6 is set to 0, and the code amount Ln (7) generated by # 7 is set. Set a sufficiently large value for. Also, in # 6, 1 is assigned to the variable FLAG used to determine whether or not all the difference values are 0. In step S46, the maximum value (max) and the minimum value (min) of the pixel values in the block are obtained, and when max-min is smaller than 2e + 1, the processing corresponding to the # 7 encoding method (after step S56) Branch to, otherwise #
Processing corresponding to the sixth encoding method (after step S47)
Proceed to.

In step S47, an approximate plane is obtained from the pixel values in the block. From step S48 to step S52, the pixel values in the block are processed in order. In step S48, the pixel value of interest X in the block and the position of each pixel of interest on the approximate plane obtained in step S47 are found. The difference value D = X−P is obtained from the value P to be calculated, and step S42
The modified difference value D ′ is calculated in the same manner as
Rewrite with + P. In step S49, the difference value between the rewritten corrected pixel value X ′ and the predicted value P is Huffman-coded by the Huffman table for plane coding, and the code length is added to Ln (6).

In step S50, it is determined whether or not the difference value between the corrected pixel value X'and the predicted value P is 0. If it is not 0, 0 is assigned to FLAG in step S51. In step S52, the end of the block is determined. If it is not the end, the process is continued from S48 for the next pixel in the block. Step S53 is a step of determining the value of FLAG. If it is 0, the process proceeds to step S54, and if not, the process proceeds to step S55. In step S54, the code amount of a, b, c0 plus one bit for FLAG is added to Ln (6). In step S55, FLAG is added to the code amounts of a, b, and c0.
The value added with 1 bit for is substituted as a new Ln (6) value.

If it is determined in step S46 that the # 7 coding method can be used, in step S56, the quotient of (max-min) / 2 is set as c0, and c0 is coded. Substituting the code amount at that time into Ln (7),
Substitute a sufficiently large value for Ln (6).

In step S64, all pixel values in the block are rewritten to c0. By these processes, # 6 and # 7
Generated code amount Ln (6), L when the No. coding method is used
n (7) can be obtained.

Next, steps S57 to S60 will be described in detail. Step S57 is a step of determining whether or not the # 8 coding method can be used. The pixel value in the block is checked, and if the appearing pixel value is two values, the step S58, If not, the process proceeds to step S50.

In step S50, a sufficiently large value is substituted for Ln (8). In step S58, of the two pixel values in the block, the one having the higher luminance value is white and the one having the lower luminance value is black, and the scan is performed in the order shown in FIG. 5 to convert the run-length data into white and black. In step S59, the code amount of the code for representing the two pixel values that appear and the code amount when the above-described run length data is Huffman-coded using the # 8 dedicated Huffman table prepared in advance are set. In addition, the generated code amount Ln (8) by # 8 is calculated.

By the processing described above, the generated code amount Ln (0) to the coding methods # 0 to # 8 ...
When Ln (8) is obtained, in step S61, the minimum code amount that is the minimum value is obtained from these, and this minimum code amount
The code amount of the block is added with the code amount of the code indicating the encoding method (the code amount LA of FIG. 7 or the total code amount LB.
Of one of the four code amounts that make up L, or the code amount L)
And the information indicating the coding method number (i) is also stored.

In step S62, the imaginary block generated when the coding method of number i is used is saved. This fictitious block is used when the block to be coded is determined (rewritten with the original pixel value) in steps S28 and S34, and finally used for rewriting the block data in step S6.

In step S63, the number i and the code amount 1 are output, and the process of selecting the optimum coding method (corresponding to S21, S25 and S32 in FIG. 7) is completed. With the above, steps S21, S25, S
This completes the description of the selection of the optimum coding method at 32.

By the above processing, the block division information and the optimum coding method in step S3 are determined.

FIG. 9 is a diagram showing an example of a block division method and an encoding method assigned to each of the divided blocks to be encoded. Reference information based on this FIG. 5 is shown in FIG. This is the end of the detailed description of S3.

Next, in step S4, the reference information obtained in S3 is stored in the memory.

In step S5, with respect to the coding method number i selected for each coding target block, the coding number counter prepared for each block size causes the number of appearances of the coding methods # 0 to # 8. And the number of appearances of the modified difference value D ′ when coded by that method is counted from # 0 to # 6, the counter corresponding to the coding method (if # 0 to # 5, It is added to the predictive coding counter corresponding to each block size, or the plane coding counter if # 6).

In step S6, step S3 (S28,
Based on S34. ), The pixel value of the block extracted in S2 is rewritten using the data of the encoding target block rewritten in the fictitious block.

In step S7, the end of the image is determined,
If it is not the end block of the image, the process from step S2 is performed on the next block, and if it is the end block, step S2.
Processing is transferred to 8.

Here, the loop processing from S2 to S7 is processing for rewriting block data (imaginary block) in block units to obtain image data. Therefore, the reference pixels (A, B, C, or
If (c0, etc.) is outside the area of the block of 16 × 16 pixels extracted in S2, the pixel value of the block data rewritten in S6 by the previous loop processing is used.

In step S8, a 16 × 16 pixel predictive coding counter, an 8 × 8 pixel predictive coding counter, 4
A prediction coding Huffman table of 16x16 pixels, a prediction coding Huffman table of 8x8 pixels, based on the values of the prediction coding counter of x4 pixels and the plane coding counter,
A 4 × 4 pixel prediction coding Huffman table and plane coding Huffman table are reconstructed. Also, the number of appearances of the optimal coding methods # 0 to # 8 is counted by the coding method number counter prepared for each block size, and the coding method number is coded based on the statistics of this count. Build a Huffman table for you.

In step S9, the optimization processing loop (S0
(S8) to S8), the process branches to step S9 if the preset number of loops have been completed, and to step S0 otherwise. In this case, S0 of the optimization processing loop of the first time and S0 of the optimization processing loop of the second and subsequent times are different in processing, and S0 of the second and subsequent times is the same as the previous S8.
The optimal Huffman table created in step 1 is used by substituting it into the reference Huffman table.

In step S10, the stored reference information is referred to and the image data rewritten in step S6 is actually encoded. FIG. 12 shows the flow of the encoding process in step S10. In FIG. 12, S80 is a step of encoding a Huffman table, S81 is a step of extracting block data of the maximum size (16 × 16 pixels) from an image, S82 is a step of extracting reference information corresponding to this block data, and S83 is a reference. The block data is encoded according to the information, and S84 is the step of determining the end of the image. A detailed description will be given below.

First, in step S80, a Huffman table for coding numbers of 16 × 16 pixels, a Huffman table for coding numbers of 8 × 8 pixels, a Huffman table for coding numbers of 4 × 4, and a prediction of 16 × 16 pixels. Huffman table for encoding,
For the prediction coding Huffman table of 8 × 8 pixels, the 4 × 4 prediction coding Huffman table, and the plane coding Huffman table, information indicating the structure of each Huffman table is output as Huffman table information.

In step S81, the image data is
Retrieve block data with maximum block size. In step S82, the reference information for the block data is retrieved from the memory.

In step S83, the reference information and the data obtained by encoding the block data having the maximum block size according to the reference information are output as an encoded data string.

In step S84, the end of the image is determined, and if there is an unprocessed block, the process is performed from step S81 on the next block.

By the above processing, a coded data string for image data is generated and output.

This is the end of the description of the first embodiment.

(Other Embodiments) The present invention is not limited to the above embodiments. For example, in the above-described embodiment, the maximum block size is set to 16 × 16 pixels to divide the blocks, but a larger block size such as 128 × 128 pixels may be set.

Further, as the method of dividing the block, the method of dividing the block into four parts having the same size has been described, but the block size may be variable.

Further, in the first embodiment, as shown in FIG. 7, the block is divided into blocks to determine the block to be encoded based on the code amount of the block to be processed and the code amount of a plurality of sub blocks obtained by dividing the block into four. However, the maximum block (16 × 16
It is also possible to obtain the code amount for all blocks from the pixel) to the minimum block (4 × 4) and then divide the blocks based on all the code amounts.

Further, first, the code amount of the minimum block is obtained, this block is regarded as a sub block, and as shown in FIG. 6, the total code amount of four sub blocks (4 × 4 pixels) and the block of the area where these are combined. The temporary block division method and the code amount are determined based on the code amount of (8 × 8 pixels), and the total code amount of the four determined regions (8 × 8 pixels) and the determined region (8 × 8 pixels). ), A new temporary block division method is determined based on the code amount of a larger block (16 × 16 pixels), and the block to be encoded is determined by repeating this up to the maximum block size. Is also good.

Although the predictive coding method, the plane coding method and the run length coding method are used as the selectable block coding method, the application method is changed by changing the prediction method, or the Markov model code is used. It goes without saying that another encoding method such as an encoding method may be used. Furthermore, although Huffman coding is used as the entropy coding method in each coding method in the above-described embodiment,
Of course, this may use other entropy coding methods such as arithmetic coding.

Further, in S6, the value of the prediction error (generated in S3) at the time of encoding the image in which the block data is rewritten may be stored without rewriting the block data. In this case, in S10, instead of re-encoding using the rewritten block data, encoding may be performed using the value of the prediction error (correction difference value) already obtained in S3. However, when the difference value corresponding to each pixel is stored, for example, the process of performing predictive coding by referring to the plurality of pixel values (A, B, C) in S41 of FIG. The circuit configuration and processing procedure for generating are complicated. Therefore, in the first embodiment, the method of storing the corrected pixel value is used instead of the method of storing the corrected difference value.

Further, FIG. 13 shows an example of a circuit configuration for performing the encoding process of the first embodiment.
Is a CPU control unit capable of performing encoding and decoding of image data by software processing, 2 is an image input unit for taking in the image data, 3 is a printer unit for printing the image data, and 4 is a display for displaying the image data. Reference numeral 5 denotes a communication control unit capable of transmitting data encoded by the CPU control unit 1 or receiving encoded data transmitted from an external device. Reference numeral 6 denotes an image memory for storing image information.

[0089]

As described above, according to the present invention, by converting pixel values or blocks in an image, a relatively simple circuit configuration or processing procedure is used, and image data can be compressed with high efficiency. It is possible to perform encoding in which the error of the pixel value is controlled.

Further, multi-valued data can be encoded with high compression efficiency and by controlling the error in pixel value.

[Brief description of drawings]

FIG. 1 is a diagram showing a flow of processing according to a first embodiment of the present invention.

FIG. 2 is a diagram showing types of encoding methods and positional relationships of pixels.

FIG. 3 is a diagram showing a portion of image data in encoded data.

FIG. 4 is a diagram explaining a Huffman code table for encoding c0 ′.

FIG. 5 is a diagram showing a method of scanning a run length in the case of # 8 encoding method.

FIG. 6 is a diagram showing how a block to be processed is divided into sub-blocks.

FIG. 7 is a diagram illustrating the flow of processing in S3.

FIG. 8 is a diagram illustrating the flow of processing in S21, S25, and S32.

FIG. 9 is a diagram showing an example of a maximum block division method and an encoded data form.

FIG. 10 is a diagram showing how a difference value is converted into a modified difference value.

FIG. 11 is a diagram illustrating a flow of encoding processing in S10.

FIG. 12 is a diagram showing an example of a circuit configuration for realizing the present invention.

FIG. 13 is a diagram showing a circuit configuration for realizing the first embodiment.

[Explanation of Codes] S0 Step of setting reference Huffman table S1 Step of reading image data S2 Step of extracting block data from image data with maximum block size S3 Step of determining block division method and encoding method S4 Saving reference information Step S5 Step for obtaining statistics about data to be encoded S6 Step for rewriting image data with a fictitious block S7 Step for determining end of image S8 Step for generating optimal Huffman table S9 Step for determining loop end S10 Actually according to reference information Steps for encoding an image

Claims (20)

[Claims] 1. An image processing apparatus for predictively encoding an image, comprising: a predicting unit that predicts a pixel value of the image; And an encoding unit that encodes a parameter relating to the image based on the result of the prediction performed by the prediction unit and the image converted by the image conversion unit. Image processing device. 2. The image processing apparatus according to claim 1, wherein there are a plurality of prediction methods by the prediction means. 3. The image-related parameter is a difference value between a pixel value of the image converted by the image conversion unit and a prediction value predicted by the prediction unit. Image processing device. 4. The image processing apparatus according to claim 1, wherein the pixel value of the image is represented by multiple values. 5. The image processing apparatus according to claim 1, wherein the encoding by the encoding unit performs different processing for each block of an image. 6. The conversion by the image conversion means is performed with an allowable error of a pixel value being ± e.
An image processing apparatus according to claim 1. 7. An image processing method for performing predictive encoding of an image, comprising: a predicting step of predicting a pixel value of the image; A conversion step of converting in a range, an image converted in the conversion step, and an encoding step of encoding a parameter relating to the image based on a result of prediction in the prediction step. Image processing method. 8. An image processing apparatus for predictively coding an image for each block image, generating means for generating a prediction parameter corresponding to the block image, and prediction error of the block image according to the prediction parameter generated by the generating means. Has a conversion means for converting the block image so that it has a predetermined value, and an encoding means for encoding the block image converted by the conversion means using the prediction parameter generated by the generation means. An image processing device characterized by: 9. The image processing apparatus according to claim 8, wherein the prediction parameter generated by the generation unit exists corresponding to each pixel value in the block. 10. An image processing method for predictively encoding an image for each block image, the generating step of generating a prediction parameter corresponding to the block image, and the prediction error of the block image according to the prediction parameter generated in the generating step. Has a conversion step of converting the block image so as to have a predetermined value, and an encoding step of encoding the block image converted in the conversion step using the prediction parameter generated in the generation step. An image processing method characterized by: 11. An image processing apparatus for encoding an image using a differential encoding method, a limiting means for limiting a value of a difference value, a predicting means for generating a predicted value for a pixel value of the image, A conversion unit that converts the pixel value of the image based on the value limited by the restriction unit and the prediction value generated by the prediction unit, the pixel value converted by the conversion unit, and the prediction generated by the prediction unit An image processing device, comprising: an encoding unit that encodes a difference value from the value. 12. The image processing apparatus according to claim 11, wherein there are a plurality of types of the differential encoding method. 13. The difference value limited by the limiting means is plural, and the difference value is set such that the plural difference values are discrete integers.
2. The image processing device according to 1. 14. An image processing method capable of differentially encoding an image, a limiting step of limiting a value of a difference value, a generating step of generating a predicted value for a pixel value of the image, and the limiting step. A conversion step of converting the pixel value of the image based on the value limited by the above and the prediction value generated in the generation step, the pixel value converted in the conversion step, and the prediction value generated in the prediction step And an encoding step of encoding a difference value between the image processing method and the image processing method. 15. Input means for inputting multi-valued data, output means for outputting a difference value using the multi-valued data input by said input means, and a difference value output by said output means is defined as an allowable error. And a control means for controlling the difference value so that the values are discretely arranged at intervals corresponding to a predetermined value, and an encoding means for encoding the difference value controlled by the control means. A characteristic image processing device. 16. The image processing apparatus according to claim 15, wherein the multi-valued data is multi-valued image data. 17. The image processing apparatus according to claim 15, wherein the difference value is a difference value between a predicted value predicted from the multivalued data and a value of the multivalued data. 18. The image processing apparatus according to claim 15, wherein the control unit controls the difference values to be discretely arranged. 19. The image processing according to claim 15, wherein the encoding by the encoding means uses entropy encoding of the difference value based on statistics of the difference value controlled by the control means. apparatus. 20. An input step of inputting multi-valued data, an output step of outputting a difference value using the multi-valued data input in the input step, and a difference value output in the output step being an allowable error. It has a control step of controlling the difference value so that the values are discretely arranged at intervals within a corresponding predetermined value, and an encoding step of encoding the difference value controlled in the control step. Characterized image processing method.