JPH05316359A - Picture data coder - Google Patents

Abstract

PURPOSE:To control the coding condition in response to kinds of pictures. CONSTITUTION:A linear quantization section 140 quantizes a 2-dimension DCT coefficient of each block outputted from a 2-dimension DC transformation section 120 based on a quantization matrix for valid coefficient count, and a variable length coding section 150 codes the quantization coefficient obtained by quantization by using a code table, for valid coefficient count. A valid coefficient count section 240 reads the coded data from a code buffer 230 to count frequencies of appearance of the valid coefficients of each block, and a slow change area discrimination section 250 obtains a ratio of a slow change area occupied in the entire picture based on the count result. A control section 260 selects a coding quantization matrix and a code table based on the ratio and the frequencies of appearance of the valid coefficients in the entire picture counted by the valid coefficient count section 240.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data coding apparatus for coding a multivalued image with high efficiency, and in particular, dividing a multivalued image into a plurality of blocks each composed of a plurality of pixels, The present invention relates to an image data encoding device that performs, after each block, multi-valued data of a plurality of pixels in that block by orthogonal transformation and then performs encoding.

[0002]

2. Description of the Related Art CCITT (International Telegraph and Telephone Advisory Committee) is promoting international standardization of high-efficiency color still image coding.
Which is a joint organization of ISO and ISO (International Standards Organization)
(Joi-nt Phtographic Experts Groug) has announced high efficiency coding using two-dimensional DCT (two-dimensional discrete cosine transform) as its recommendation.

A high-efficiency coding system for color still images recommended by JPEG (hereinafter referred to as JPEG system)
Divides the input multi-valued image into blocks of 8 × 8 pixels, and the image signal (multi-valued data) of 8 × 8 pixels in each block is spatially transformed by two-dimensional discrete cosine transform (hereinafter referred to as two-dimensional DCT) Converted to 8 × 8 two-dimensional DCT coefficients of frequency distribution, then quantized these two-dimensional DCT coefficients with a threshold value adapted to the visual sense, and statistically obtained the quantized coefficients obtained by the quantization. Variable length coding is performed by the Huffman table.

FIG. 16 is a basic block diagram of a conventional image data coding apparatus for coding a multivalued image by the JPEG system. In the image data encoding device shown in the figure, a multi-valued image is divided into a plurality of blocks of 8 × 8 pixels, and is input from the terminal 110 to the two-dimensional DCT conversion unit 120 in block units. The two-dimensional DCT transform unit 120 orthogonally transforms an input image signal of 8 × 8 pixels by a two-dimensional DCT transform to obtain 8 × 8 two-dimensional DC of spatial frequency distribution.
It is converted into a T coefficient and output to the linear quantization unit 130.

The linear quantizer 130 receives the input 8 × 8.
Each of the two-dimensional DCT coefficients is linearly quantized using the corresponding quantization threshold in the quantization matrix 140. As a result of this quantization, the two-dimensional DCT coefficient having a value smaller than the corresponding quantization threshold in the quantization matrix 140 becomes “0”, and only the DC component (DC coefficient) and a few AC component (AC coefficient) are “0”. A quantized coefficient having a value other than “” is generated.

The quantized coefficient quantized by the linear quantizer 130 as described above is input to the variable length encoder 150. The variable length coding unit 150 refers to the variable length coding table 160 configured by the Huffman table separately provided for each of the DC component and the AC component, and performs variable length coding on the sequence of the input quantization coefficient. To do. That is, for the DC component, the difference value between the DC component at the beginning of each block and the DC component of the preceding block is variable length coded by referring to the one-dimensional Huffman table. For the AC component, the combination of the value (index) of the effective coefficient whose value is not “0” and the run length (run length) of the invalid coefficient whose value up to the effective coefficient is “0” is variable. Long code.
In this way, the variable-length coding unit 150 receives the DC component, A
Each of the C components is subjected to variable length coding using a code table 160 composed of a Huffman table, and code data obtained by the coding is output from a terminal 170.

In such encoding of a multi-valued image by the JPEG method, the quantization coefficient is determined by the value of the two-dimensional DCT coefficient and the corresponding quantization threshold value in the quantization matrix 140. Then, the above-described processing is performed on all blocks of one screen in block units, so that the two-dimensional DCT coefficient for one screen is quantized.

[0008]

As described above, in the conventional JPEG system, when the two-dimensional DCT coefficients of each block are quantized in block units, the two-dimensional DCT coefficients of all blocks are quantized by one type of quantum. It is quantized by the quantization matrix. That is, 2 of each spatial frequency in each block
The dimensional DCT coefficient is quantized with the same quantization threshold in all blocks.

By the way, in general, for human vision, with respect to an image (for example, a landscape image) whose gradation changes drastically, even if the restored image has some approximation error with the original image, the image quality is deteriorated. It tends to be difficult to recognize. On the other hand, for an image (for example, a portrait) in which gradation changes gently, human vision tends to recognize the approximation error of the restored image with respect to the original image, and easily recognize image quality deterioration. is there.

Further, in human vision, the evaluation of the image quality of the entire image tends to be dominated by a part of the image quality (the worst image quality part). That is, for example, even in an image with a sharp gradation change like a grove of landscape paintings, human vision is
For example, when the blue sky portion where the gradation gradually changes occupies about 1/3 of the entire image, the image quality of the entire image tends to be evaluated by the image quality of the blue sky.

The approximation error is caused by the loss of some of the two-dimensional DCT coefficients when quantizing the two-dimensional DCT coefficients. Therefore, the two-dimensional DC in the block
When the T coefficient is quantized with one kind of quantization threshold value (quantization matrix), for example, there is a problem that image quality deterioration is conspicuous in an image in which gradation is gradually changed. On the other hand, in order to prevent this, if each of the quantization thresholds in the quantization matrix is set to a low value, the amount of code increases when encoding an image with sharp gradation changes, and efficient data compression cannot be achieved. It had the drawback of creating new problems.

In order to solve the above-mentioned conventional problems, the present invention is capable of encoding all images having different gradation changes while ensuring high image quality of a restored image. It is an object to provide an encoding device.

[0013]

FIG. 1 is a principle block diagram (No. 1) of the present invention (first invention). The present invention is
After dividing the original image into multiple blocks consisting of multiple pixels, the tone values of multiple pixels in these multiple blocks are orthogonally transformed for each block, and the orthogonal transformation coefficients obtained by the orthogonal transformation are calculated. It is premised on an image data encoding device that quantizes and encodes the quantized coefficient obtained by the quantization.

The effective coefficient counting means 1 counts the number of appearances of the quantized effective coefficient for each of the plurality of blocks. The effective coefficient is an orthogonal transform coefficient that is not "0" among the orthogonal transform coefficients of the AC component.

The coding condition setting means 2 sets a coding condition based on the number of appearances of the effective coefficients for each of the blocks counted by the effective coefficient counting means 1. The encoding unit 3 encodes the original image by the above method (orthogonal transformation / quantization / encoding) using the encoding condition set by the encoding condition setting unit 2.

In the first invention, the coding condition setting means 2 includes, for example, as described in claim 2, a detection means 2a for detecting a block in which the number of appearances of the effective coefficient is smaller than a predetermined value. Judgment means 2 for judging whether or not the detection result of the detection means 2a satisfies a predetermined condition.
b and a setting means 2c for setting the above-mentioned coding condition based on the judgment result of the judgment means 2b (see FIG. 2).

In the first aspect of the invention, the effective coefficient counting means 1 may be configured to count the number of blocks in which the number of appearances of the effective coefficient is smaller than a predetermined value. Of the original image, and the second counting means 1b that counts the number of appearances of the effective coefficient in all blocks of the original image, the encoding condition setting means 2 includes the first counting means 1a. 1a counting result and second
The encoding condition may be set based on the counting result of the counting means 1b (see FIG. 3).

Next, FIG. 4 is a principle block diagram of a second invention according to claim 4. This second invention is also the first invention described above.
The original image is divided into a plurality of blocks composed of a plurality of pixels in the same manner as in the above invention, the gradation values of a plurality of pixels in the plurality of blocks are orthogonally transformed for each block, and then the orthogonal transformation is performed. It is premised on an image data encoding device that quantizes the obtained orthogonal transform coefficient and encodes the quantized coefficient obtained by the quantization.

The effective coefficient counting means 11 counts the number of appearances of the quantized effective coefficient for each of the plurality of blocks. The block number counting unit 12 counts the number of consecutive blocks in which the number of appearances of the effective coefficient is smaller than a predetermined value, based on the counting result of the effective coefficient counting unit 11.

The coding condition setting means 13 sets a coding condition based on the counting result of the block number counting means 12. The encoding means 14 encodes the original image by the above method (orthogonal transformation / quantization / encoding) using the encoding condition set by the encoding condition setting means 13.

In the second aspect of the invention, for example, as described in claim 5, the effective counting means 11 is the first counting means 11a for counting the number of blocks in which the number of appearances of the effective coefficient is smaller than a predetermined value. And a second counting means 11b for counting the number of appearances of the effective coefficient in the entire original image, the block number counting means 12 is based on the counting result of the first counting means 11a. Then, the number of blocks in which the number of appearances of the effective coefficient is smaller than a predetermined value is counted, and the coding condition setting means 13 causes the block number counting means 12
The coding condition may be set based on the counting result of 1) and the counting result of the second counting means 11b.

[0022]

According to the first aspect of the invention, the original image is divided into a plurality of blocks composed of a plurality of pixels (for example, 8 × 8 pixels), and the gradation value of each pixel in each block is orthogonally transformed in block units. To be done. Then, the orthogonal transformation coefficient of each block obtained by the orthogonal transformation is input to the effective coefficient counting means 1.

The effective coefficient counting means 1 is an effective coefficient (A which is not "0") out of the orthogonal transformation coefficients in each input block.
The number of C component orthogonal transform coefficients) is counted for each block. The encoding condition setting means 2 compares the effective coefficient of each block input from the effective coefficient counting means 1 with a predetermined value, and the number of effective coefficient is less than the predetermined value. To detect.

Next, the coding condition setting means 2 judges whether or not the detected number of blocks is equal to or larger than a predetermined threshold value, and based on the judgment result, the coding condition (quantization) is determined. The quantization matrix and the code table for variable-length coding used when performing the setting are set.

The encoding means 3 inputs the orthogonal transformation coefficient of each block of the original image again, and the orthogonal transformation coefficient of each block is, for example, JP according to the above-mentioned encoding condition.
Variable length coding is performed in accordance with the EG method to create coded data of the original image.

A block in which the number of appearances of the effective coefficient is small indicates a region in which the gradation change is gradual. Therefore, when such coding is performed, the number of blocks in which the number of the appearances of the effective coefficient is small is counted, It is possible to determine the ratio of the region in which the tonal change is gentle to the entire original image. Therefore, by controlling the encoding condition according to the type of the image, it is possible to perform encoding capable of restoring the original image with high image quality while suppressing an increase in the code amount.

Further, as described in claim 3, in addition to the detection of the block in which the gradation changes gradually, the effective coefficient counting means 1 also counts the number of appearances of the effective coefficient in the entire original image, so that the encoding condition is satisfied. The setting means 2 can also judge the complexity of the entire image, and by controlling the coding condition by further considering this judgment, more effective high efficiency coding becomes possible.

Next, in the second invention, the effective coefficient counting means 11 counts the number of appearances of the effective coefficient for each block, and the block number counting means 12 determines the number of appearances of the effective coefficient, for example. The counting result of the effective coefficient counting means 11 is sequentially input, and the number of consecutive blocks in which the number of appearances of the effective coefficient is smaller than a predetermined value is counted.

The coding condition setting means 13 is based on the counting result of the block number counting means 12, and the coding conditions when the coding means 14 codes the original image (quantization matrix for quantization and variable length). Code table for encoding) is set.

In this case, the counting result of the block number counting means 12 serves as a standard for knowing the size of the flat portion where the gradation change is small. Therefore, the coding condition setting means 13 judges the size of the flat area in the original image in which the gradation change is gentle from the counting result of the block number counting means 12, and the coding means 14 variable-length codes the original image. It is possible to control the coding condition used when performing.

Therefore, also in the second aspect of the invention, the encoding condition is controlled on the basis of the ratio of the region of the original image in which the gradation change is gentle to suppress the increase in the code amount and restore the original image with high image quality. Possible encoding can be done.

Further, as described in claim 5, the effective coefficient counting means 11 counts the number of appearances of effective coefficients in the entire original image by the second counting means 11b, so that the encoding condition setting means 13 In addition to the number of blocks described above, if the coding conditions are set in consideration of the complexity of the entire image based on the number of appearances of effective coefficients in the entire original image, more effective encoding of the original image can be performed. You can

[0033]

Embodiments of the present invention will be described below with reference to the drawings. Two-dimensional DCT such as the above JPEG method
In the image data encoding method using, when all the two-dimensional DCT coefficients in each block are quantized with a constant quantization threshold value regardless of the image type after the two-dimensional DCT conversion,
The number of two-dimensional DCT coefficients that remain as effective coefficients (area coefficients having a coefficient value other than “0”) depends on the intensity of the gradation change within the block (the magnitude of the change in the gradation value of the pixel or the change in the gradation value). The size of the cycle, etc.) changes.

In this embodiment, the number of effective coefficients obtained when all the two-dimensional DCT coefficients in a block are quantized with a constant quantization threshold value and the intensity of gradation change in the block. Use the correspondence relationship with. That is, the number of appearances of the effective coefficient is counted, and the complexity of the entire image is determined based on the counting result, and the ratio of the region in which the gradation change is gentle to the entire image is determined. It is determined whether the image has a sharp tone change or an image that includes many flat gradations, and the encoding condition of the image data is controlled based on the determination result to increase the code amount. While suppressing, secure a high restored image.

FIG. 6 is a basic block diagram of an image data coding apparatus according to an embodiment of the present invention. In the figure,
With respect to the same block as the block in the image data encoding device shown in FIG. 16 described above, the same name and the same code contribute, and detailed description thereof will be omitted.

The quantization matrix storage unit 210, which is not particularly shown, is a quantization matrix A used by the linear quantization unit 140 to create a quantized coefficient for effective coefficient counting, and also the linear quantization unit 140. Stores a plurality of quantization matrices B for encoding image data used for linearly quantizing the two-dimensional DCT coefficients of each block input from the two-dimensional DCT transform 120.

Although not shown in the figure, the code table storage unit 220 includes a code table A used by the variable length coding unit 150 to create data for counting effective coefficients, and a variable length coding unit 150. A variable-length code table (Huffman code table) used when variable-length coding is performed by compressing the quantized coefficient of each block input from the linear quantization unit 140 is stored.

The code buffer 230 includes a variable length coding unit 1.
This is a buffer for storing code data for counting effective coefficients for one image output by 50 or variable length code data for the original image output by the variable length coding unit 150.

The effective coefficient counting section 240 reads the code data for counting the effective coefficient from the code buffer 230, and outputs the appearance number C of the effective coefficient in the entire image to the control section 260. Further, the effective coefficient counting unit 240 obtains the number of appearances of the effective coefficient for each block from the coded data for counting the effective coefficient, and based on the number of appearances of the effective coefficient, a region where the gradation change is gentle for each block. (Hereinafter, referred to as a gradual region for convenience), and the determination result is output to the gradual region determination block 250.

The gradual area determination unit 250 sends the number of flat portion blocks B, which is the number of blocks with gradual gradation change in the original image, to the control unit 260 based on the determination result input from the effective coefficient counting unit 240. Output.

The control unit 260 stores the quantization matrix on the basis of the appearance number C of the effective coefficients in the entire original image input from the effective coefficient counting unit 240 and the flat portion block number B input from the gradual area determination unit 250. The optimum quantization matrix is output from the plurality of variable-length coding quantization matrices B stored in the unit 210 to the linear quantization unit 140, and the code-table storage unit 220 outputs the variable-length coding quantization matrix B for variable-length coding. The Huffman code table is output to the variable length code variable length unit 150.

Next, the basic operation of the image data coding apparatus having the above configuration will be described with reference to the flowchart of FIG. First, the control unit 260 controls the gradual area determination unit 2
After initializing 50 (SA1), the quantization matrix A for quantization of the effective coefficient is calculated from the quantization matrix storage unit 210.
Is output to the linear quantization unit 140, and at the same time, the code table storage unit 220 outputs the code table A for creating code data for counting effective coefficients to the variable length coding unit 150 (SA.
2).

Subsequently, the two-dimensional DCT transforming unit 120 starts a two-dimensional DCT in block units from the head block of the original image.
The linear quantization unit 140 performs the transformation, and the two-dimensional DCT
The two-dimensional DCT coefficient in each block obtained by the transform is linearly quantized using the quantization matrix A. And
Next, the variable length coding unit 150 creates code data for counting effective coefficients from the quantized coefficients obtained by the linear quantization with reference to the above code table A, and calculates code data for counting effective coefficients of each block. The process of storing the code data in the code buffer 230 is performed for all blocks (SA1, SA2).

Subsequently, the effective coefficient counting unit 240 reads out the code data for counting the effective coefficients of all the blocks from the code buffer 230 in a predetermined order in block units, and the number of appearances C of the effective coefficients in all the blocks is C. In addition, it is determined for each block, based on the number of appearances NB of effective coefficients in the block, whether or not the block is a block with a gradual gradation change, and the determination result is a gradual area determination unit. It is output to 250 (SC1).

The gradual area determination unit 250 determines the number B (flat portion blocks) of the blocks (flat portion blocks) whose gradation changes gently based on the determination result of the gradual area for each block input from the effective coefficient counting unit 240. The number of partial blocks B) is counted and the number of flat portion blocks B is output to the control unit 260.

The control unit 260 determines the property (type) of the image based on the number C of appearances of effective coefficients in all the blocks input from the effective coefficient counting unit 240 and the number B of flat portion blocks input from the gradual area determination unit 250. ) Is determined (SC2),
From the quantization matrix storage unit 210, the quantization matrix B to be used in the linear quantization in the variable length coding process.
To the linear quantization unit 140, and also to the variable length coding unit 150 from the code table storage unit 220 to output the variable length coding table (Huffman coding table) used when variable length coding the quantized coefficient. (SC3).

As described above, after selecting the quantization matrix for linear quantization and the variable length code table for variable length coding the quantized coefficient, the quantization matrix and variable length code table are selected. The two-dimensional DCT transform unit 120 is used in the same manner as the conventional image data coding device described above.
The linear-quantization unit 140 and the variable-length coding unit 150 perform variable-length coding on the original image, and the variable-length code data obtained by the variable-length coding is output from the terminal 170 to the outside via the code buffer 230. Output (SD1, SD2
2).

Next, FIG. 8A shows a configuration example of the effective coefficient counting section 240. In FIG. 9A, the intra-block effective counter 241 reads the effective coefficient counting code data of the format shown in FIG. 9 from the beginning from the code buffer 230 shown in FIG. 6, and calculates the effective count of each block. The number of appearances NB is output to the comparison unit 243.

The code data for counting the effective coefficient is:
A bit of "0" indicates the end of the block EOB (End Of
Blook) signal, and each 1 bit of the bit string of "1" between EOB signals corresponds to one effective coefficient in the block. Therefore, the bit of "0" (EOB
The number of bits of "1" between signals is equal to the number NB of appearance of effective coefficients in a block. In addition, the bit of "0" (E
When j (j = 2, 3, ...) Consecutive OB signals are present, it means that there are no effective coefficients in consecutive (i−1) blocks. Thus, the code data for counting the effective coefficient is a very compact code.

An example of the structure of the intra-block effective coefficient counting unit 241 is shown in FIG. In the figure, the "1" counting unit 241a sequentially reads the code data for valid coefficient counting in the format shown in FIG. 9 in the code buffer 230 from the first bit until the bit "0" (EOB signal) is detected. The number of consecutive 1 bits NB ₁ is counted. And
When the "1" counting unit 241a detects the EOB signal (bit "0"), the counting value NB ₁ (equal to the number of appearance of the effective coefficient of the first block) up to that point is stored in the cumulative addition storage unit 241a. In addition, the comparison unit 2 shown in FIG.
It also outputs to 43. The "1" counting unit 241a continues to "1" until the EOB signal (bit "0") is detected again from the next bit (head of the second block) of the EOB signal of the code data for counting the effective coefficient. When the EOB signal (“0” bit) is detected by counting the number of consecutive “NB ₂ ” bits, the number of consecutive “1” bits (the number of appearance of effective coefficients in the second block) NB ₂ until then is calculated. , The number NB ₁ stored in the cumulative addition storage unit 241b, and the addition result (NB ₁ + NB ₂ ) is added to the cumulative addition storage unit 241.
In addition to storing in b, the counting result NB ₂ is output to the comparing section 243. Thereafter, in the same manner, the “1” counting unit 2
The reference numeral 41a reads the effective coefficient counting data, and the effective coefficient numbers NB ₃ and NB are also applied to the third and subsequent blocks.
_4, by counting the ..., and outputs the count result to the comparator unit 243, will store the accumulated value of the effective coefficient accumulating storage unit 241b. Therefore, the “1” counting unit 24
When 1a has read out all the code data for counting effective coefficients, the cumulative addition section storage section 241b stores the number C of appearances of effective coefficients in the entire original image. Then, the appearance number C of the effective coefficient in the entire original image is read by the control unit 260.

Returning to FIG. 9A again, the explanation will be made as follows.
The comparison unit 243 receives the number of effective coefficients NB _i (i = 1, 2, 3, ...) Of each block from the in-block effective coefficient counting unit 241.
..) are input in order from the first block, and the number NB _i
(I = 1, 2, 3 ...) are sequentially stored in the threshold value Th _n (eg,
"1") as compared to the determination result of the if NB ≦ Th _n "0", and outputs a determination result if NB> Th _n "1" to the gentle area determination unit 250.

Here, FIG. 10 shows an example of the configuration of the gradual area determination section 250. In the figure, the block comparison result storage unit 251 determines whether or not the number of appearances of effective coefficients for all blocks of the original image sent from the effective coefficient counting unit 240 is small (“0”) (“1”). The judgment result of is stored.

The adder / subtractor 252 reads the determination results of each block from the block comparison result storage 251 in order from the first block, performs addition and subtraction on the determination results of two consecutive blocks, and adds and subtracts them. Only when the results are the same, the continuous number addition storage unit 25
The value of 3 is added by "1".

Both the addition result and the subtraction result are equal to each other when the determination results of the two consecutive blocks are both "0", that is, the two consecutive blocks are flat blocks with little gradation change. Only in some cases.

When the addition result and the subtraction result match in this way, the addition / subtraction unit 252 sets the latest block from which the determination result is read from the block comparison result storage unit 252 as the previous block, and then the previous block. The determination result of the block next to the block is read from the block comparison result storage unit 252, and the determination results of those blocks are again read.
The above addition / subtraction processing is performed. Then, while the addition and subtraction results match, the process of adding "1" to the continuous number addition storage unit 253 is repeated. Then, when the results of the addition and the subtraction do not match, the comparison unit 25
Start 4.

The comparison unit 254 has a continuous number addition storage unit 253.
The value NC (which is equal to the number of consecutive flat blocks) is compared with the threshold value Th _c stored in the consecutive number threshold storage unit 255, and the flat portion is _calculated only when NC ≧ Th _c. The number of consecutive flat blocks NC is added to the value of the register inside the detection unit 256. Hereinafter, the same process is repeated until the above-mentioned determination results (data indicating whether the number of effective coefficients is small) of all blocks are read from the block comparison result storage unit 251.

When the flatness detecting section 256 has finished processing the above judgment results for all blocks of the original image, the flatness detecting section 256 sets the value B stored in the internal register to the control section 26 shown in FIG.
Output to 0. This value B is equal to the number of blocks belonging to a region where the number of flat blocks having a small number of effective coefficients (gradation change is small) continues for a predetermined number Th _c or more, that is, the number B of flat portion blocks.

Next, the operation of the image data encoding apparatus having the above configuration will be described with reference to the flow charts of FIGS. 6, 8, 10 and 11. First, the control unit 26 shown in FIG.
0 is initialized, and the "1" counting unit 241a and the cumulative addition storage unit 241b in the block internal effective coefficient counting unit 241 in the effective coefficient unit 240 are initialized to "0" (see FIG. 8). Block comparison result storage unit 251 and continuous number addition storage unit 25 in the region determination unit 250
3, and the registers in the flat portion detection unit 256 are initialized to "0" (see FIG. 10).

When the control unit 260 is initialized, the three sets of quantization matrices (quantization matrix A for effective coefficient counting, 2 for linear quantization) stored in the quantization matrix storage unit 210 are initialized. Two quantization matrices B ₁ ,
B ₂ ), the quantization matrix A for effective coefficient counting is selectively output to the linear quantization unit 140, and three sets of code tables (code table for effective coefficient counting) stored in the code table storage unit 220 are selected. A and A, variable length code tables B ₁ , B ₂ ) are used to selectively output the code table A for counting effective coefficients to the variable length coding unit 150.

Quantization matrix A for counting the effective coefficients
An example of the configuration is shown in FIG. This quantization matrix A
The quantization thresholds corresponding to the 8 × 8 pixels in the block of the original image are all set to “16” uniformly.

FIG. 13 shows a structural example of the code table A for counting the effective coefficients. The code table A has a configuration in which 1-bit "1" is assigned to one effective coefficient for encoding regardless of the run value (run length).

The variable length coding unit 150 uses the code table A
The encoding is performed only with respect to the AC component (effective coefficient) representing the state of gradation change in the block by referring to
Ignore the C component. By adopting such a method, it is possible to significantly reduce the amount of coded data for counting effective coefficients to be created, as compared with the case where all two-dimensional DCT coefficients in a block are coded. Also, the variable-length coding unit 150 adds an EOB (End Of Blook) signal indicating the end of a block each time the coding of the effective coefficient by the above method is completed for each block. In this embodiment, 1-bit "0" is assigned as the EOB signal. As a result, the variable length coding unit 150 creates
The code data stored in the code buffer 250 has, for example, the format shown in FIG. 9 described above.

Then, as described above, the two-dimensional DCT
The conversion unit 120, the linear quantization unit 140, and the variable-length coding unit 150 perform coding processing for counting effective coefficients, and code data for counting effective coefficients of all blocks of the original image are stored in the code buffer 230. After the is stored,
This will be described with reference to the flowcharts of FIGS. 8, 10 and 11.

First, the in-block effective coefficient counting unit 241 of the effective coefficient counting unit 240 reads the above-mentioned effective coefficient counting code data stored in the code buffer 230 one bit at a time from the beginning, and outputs the bit value. Is a bit indicating a valid coefficient (SE1), and a "1" counter 241a is detected each time a "1" bit is detected until a "0" bit, that is, an EOB signal is detected. First, the number of appearances of the effective coefficient NB ₁ of the first block (first block) is counted (SE
2).

The "1" counting section 241a has its first
Number of appearances of effective coefficient of one block NB ₁To the comparison unit 243
It is output and stored in the cumulative addition storage unit 241b.
All effective coefficients C ("0" due to the above initialization
The number of occurrences NB₁Is added (SE3).

The comparing unit 243 determines the number of effective coefficients NB ₁ of the input first block as the effective coefficient number threshold storage unit 24.
2 is compared with the threshold value Th _n stored in 2, and NB ₁ > Th
_{If n} (SE4, YES), the gradual area determination unit 2
"1" is stored in the relevant address of 50 (SE5), while if NB ₁ ≦ Th _n (SE4, NO), the block comparison result storage unit 251 in the gradual area determination unit 250
"0" is stored in the address concerned (SE6).

The effective coefficient counting section 240 uses the process SE1.
~ SE6 is valid stored in the code buffer 230
For codes after the second block of code data for coefficient counting
Repeatedly, even if it is stored in the cumulative addition storage unit 241b
The second block for all effective coefficients C in the entire original image
Number of effective coefficients NB in each subsequent block NB₂, N
B ₃, ... are sequentially added cumulatively (SE
3), valid in each block after the second block
Number of appearances of coefficient NB₂, NB₃, ... are the effective coefficient pieces
Number threshold Th_n(SE4) and depending on the comparison result
“0” or “1” in the gradual area determination unit 250
The addresses are sequentially sorted to the corresponding addresses in the block comparison result storage unit 251.
Pay (SE5, SE6).

Then, the effective coefficient counting section 240 reads out the code data for counting the effective coefficients from the code buffer 230 for all the blocks, and the above SE1 to SE for all the blocks.
If it is determined that the process of 6 is completed (SE7, YES),
The number (appearing number) C of effective coefficients included in all blocks of the original image stored in the cumulative addition storage unit 241b is output to the control unit 260 (SE8).

Subsequently, the gradual area determination unit 250 outputs the determination results (“0” when the number of effective counts is small, “1” when the number of effective counts is small) from the block comparison result storage unit 251. , Sequentially from the first block, in the order of blocks, and the addition / subtraction unit 252 first adds and subtracts the respective determination results of the first block and the second block (SF1).

Then, the gradual area determination section 250 has a gradual change in gradation when both the addition value and the subtraction value are "0", that is, the determination results of both the first block and the second block are "0". It is determined whether the block is a flat block (S
F2), if both blocks are flat, the continuous number addition storage unit 253 is incremented by "1" (SF3). Then, the gentle area determination unit 250 returns to the processing SF1 again,
This time, the second block is set as the previous block, the determination result of the third block is read from the block comparison result storage block 251, and the determination results of the second block and the third block are added and subtracted, and the addition result and the subtraction result are both It is determined whether or not it is "0" (SF2), and if both are "0" again, this time the third block is set as the previous block and the process returns to the process SF1 again.

In this way, the continuous number addition storage unit 253
In, is stored the number of consecutive flat blocks with a gradual change in gradation. Then, the gradual area determination 250 activates the comparison unit 254 when the addition result and the subtraction result do not match in the determination processing SF2, that is, when a block having a large gradation change with many effective coefficients is read (SF2, NO). The comparison unit 254 determines that the number of consecutive flat blocks stored in the consecutive number addition storage unit 253 is the consecutive number threshold storage unit 2
The number of consecutive thresholds Th _c stored in 55 is compared, and if the number of consecutive ≧ Th _c (SF4, YE
S), the continuous number is added to the register in the flat part detection unit 256 (SF5). On the other hand, the comparison unit 254 determines that the number of consecutive times <T
If h _c is determined, the gradual area determination unit 250
Determines whether or not the process SF1 has been performed for the determination result of the previous block, and if the above SF1 has not been completed for the determination results of all blocks (SF6, NO), the determination results of the remaining blocks are compared with the block comparison result. Storage unit 25
The process is read from 1, and the processes after the process SF1 are repeated again.

Then, with respect to the determination results of all the blocks of the original image stored in the block comparison result storage unit 251, when the above processing SF1 to SF5 is completed (SF6, Y
ES), the flatness detecting unit 256 outputs the flat part block number B stored in the internal register to the control unit 260 (SF7).

The control section 260 uses the gentle area determination section 25.
When the number B of flat portion blocks is input from 0, first, the number C of appearance of effective coefficients in the entire original image, which is input from the effective coefficient counting unit 240 in the process SE8, is compared with a predetermined threshold value T2, and C ≧ threshold value. If T2 (SG1, YES), then compare the number of flat portion blocks B with a predetermined threshold value T1,
It is determined whether B ≧ threshold value T1 (SG2).

Then, if B <threshold value T1 (SG2, NO), the control unit 260 controls the quantization matrix storage unit 21.
A coarse quantization matrix (quantization condition) B1 having a large quantization threshold from 0 is selectively output to the linear quantization unit 140 (SG3), and the code table storage unit 220 sends the code table B ₁ to the variable length coding unit 150. Selective output (SE4).

On the other hand, the control unit 260 controls the discrimination processing SG.
If C <threshold value T2 in (1) (SG1,
NO), the quantization matrix storage unit 210 is caused to selectively output the fine quantization condition B2 having a small quantization threshold to the linear quantization unit 140 (SG5), and the code table storage unit 220 is also provided.
To output the code table B2 to the variable length coding unit 150 (SG6).

Further, the control unit 260 controls the discrimination processing SG described above.
2 and B ≧ threshold T1 (SG2, YES), that is, C ≧ threshold T2 and B ≧ threshold T1,
As in the case of C <threshold value T2, the quantization condition B2 and the code table B2 are set to the linear quantization unit 140 and the variable length coding unit 1, respectively.
Selectively output to 50 (SG5, SG6).

As described above, after the quantization matrix for variable length coding and the code table (Huffman code table) are selected, the two-dimensional DCT conversion section 120, the linear quantization section 140,
Also, the variable length coding unit 150 performs variable length coding of the original image using them. This variable length coding process is performed in the same manner as the conventional image data coding device described above.

As described above, first, when the number C of appearance of effective coefficients in the entire image, which is closely related to the complexity of the entire image, is smaller than a predetermined threshold value T2, the entire image is an image in which the gradation change is gentle. Therefore, it is judged that the image quality deterioration of the restored image is easily recognized, and the quantization matrix B2 having a small quantization threshold is selected, and fine quantization is performed so that the restored image is restored with a high image. To do. In this case, even if such fine quantization is performed, the number of appearances of effective coefficients is small, and the code amount is not so large.

When the number C of appearance of effective coefficients in the entire image is equal to or more than the threshold value T2, the number B of flat portion blocks indicating the size of a flat region in which gradation change is gentle is further considered. , Select the quantization matrix. That is, the number B of blocks in the flat portion is compared with a predetermined threshold value T1, and if B ≧ T1, the gradation of the original image is drastically changed over the entire image and the area of the gradation is gradually occupied over the entire image. It is determined that the ratio is also high, and in order to improve the image quality of the restored image in the gradual flat region where the gradation change is apt to be noticeable, the quantization matrix B2 having a small quantization threshold is selected, and Fine quantization is performed so that the deterioration of the image quality in the region where the tone change is gentle is not noticeable. In this case, although the number C of appearances of the effective number is large in the entire original image, since the flat region where the gradation change is large is also large, even if fine quantization is performed, the code amount does not increase so much.

Further, C ≧ T2 and B <T1, that is,
If the original image is an image in which the flat area where the gradation change is large in the entire image and the gradation change is gradual is small, the deterioration of the image quality of the restored image of the flat area is not so noticeable. Then, coarse quantization is performed by the quantization matrix B1 having a large quantization threshold.

As described above, in the present embodiment, the original image is highly efficiently encoded while maintaining the high image quality of the restored image regardless of the type of the image quality of the original image. Next, another configuration example of the gradual area determination unit 250 is shown in FIG.

The gradual area determination unit 250 sequentially accumulates the number of continuous flat blocks in the entire original image, and outputs the total number B of continuous flat blocks in the entire original image to the control unit 260.

Therefore, similar to the gradual area determination section 250 shown in FIG. 10, the block comparison result storage section 25 is used.
1 and an addition / subtraction unit 252 are provided, and in addition to these blocks, a flat block number addition storage unit 259 is added in which “1” is added each time a continuous flat block is detected by the addition / subtraction unit 252.

FIG. 15 is a flow chart showing the operation of the main part of the image data encoding apparatus using the gradual area determination section 250 according to the present invention. The operation of this image data encoding apparatus is different from the operation of the image data encoding apparatus of the first embodiment described above only in the operation of the gradual area determination unit 250, and the operations of the effective coefficient counting unit 240 and the control unit 260 are
This is similar to the image data encoding device described above.

That is, the gradual area determination unit 250 of this image data encoding apparatus determines that two continuous blocks are both flat blocks, and that the addition value and the subtraction value of the determination results of these two blocks are both It is detected when it becomes "0" (SF2, YES), and the value of the flat block number addition storage unit 259 is incremented by "1" for each detection (SF26), and these processes are performed by the block comparison result storage unit. The total number B of consecutive flat blocks in the entire original image is counted by performing the determination results of all blocks stored in the block 251 and the flat block number storage unit stores the determination result of all blocks when the above process is completed. The total number B of continuous flat blocks in the entire original image stored in 259 is output to the control unit 260 (SF7).

In the above embodiment, the number C of appearance of effective coefficients in the entire image is also used for the determination when selecting the quantization matrix (quantization condition), but the gradual area determination unit 25
The quantization matrix (quantization condition) may be selected using only the result of 0 (the number B of flat blocks in the entire image).

Further, in the above two embodiments, the continuity of the flat blocks may be detected by the area (the number of blocks) considering the continuity in the horizontal direction, the vertical direction, and both the vertical and horizontal directions. Good. These can be easily realized by changing the order of reading the block determination result by the gradual region determination unit 250.

Further, in both of the above-mentioned two embodiments, the quantization matrix (quantization condition) for variable-length coding is based on the number of flat blocks (the number of blocks forming a region where the change is gentle) in the entire original image. ) Is selected, but
The quantization matrix (quantization condition) may be selected according to the ratio of the number B of flat blocks to all blocks of the original image.

Further, in the above embodiment, one of the two types of quantization matrices (quantization threshold values) is selected, but the degree of change in the entire image (complexity of the image) is made more detailed. By making a determination, the most optimal quantization matrix may be selected from the three or more types of quantization matrices as the quantization matrix for variable length coding.

In the above embodiment, the orthogonal transform of the image data is performed by the discrete cosine transform (DCT), but in the present invention, the discrete cosine transform (DST), the Karhunen-Loeve transform, the discrete Legendre transform. ,
The present invention can also be applied to an image data encoding device that encodes image data by using other orthogonal transforms such as Hadamard transform and Harr transform.

[0091]

As described above, according to the present invention,
After dividing the original image into multiple blocks consisting of multiple pixels, the gradation values of multiple pixels in these multiple blocks are orthogonally transformed for each block, and the orthogonality of each block obtained by the orthogonal transformation The transform coefficient is quantized under a specific quantization condition, and the number of appearance of effective coefficients of each block obtained by the quantization is counted. Then, the variable length coding of the original image is performed by controlling the coding condition used when the variable length coding of the original image is performed based on the number of appearances of the effective coefficient of each block. It is possible to encode various types of images in which the degree of change and the ratio of a region having a gradual change in gradation occupy the entire image are different so that the original image can be restored with high image quality while suppressing an increase in the code amount.

[Brief description of drawings]

FIG. 1 is a block diagram (1) of the principle of the present invention.

FIG. 2 is a block diagram (No. 2) of the principle of the present invention.

FIG. 3 is a principle block diagram (No. 3) of the present invention.

FIG. 4 is a principle block diagram (No. 4) of the present invention.

FIG. 5 is a principle block diagram (5) of the present invention.

FIG. 6 is a basic block diagram of an image data encoding device according to an embodiment of the present invention.

FIG. 7 is a flowchart illustrating a basic operation of the image data encoding device according to the one embodiment.

FIG. 8 is a diagram illustrating a configuration example of an effective coefficient counting unit.

FIG. 9 is a diagram showing a format of code data for counting effective coefficients.

FIG. 10 is a diagram illustrating a configuration example of a gradual area determination unit.

FIG. 11 is a flowchart illustrating the operation of the main part of the present invention of the image data encoding device according to the above-described embodiment.

FIG. 12 is a diagram showing a configuration example of a quantization matrix A for effective coefficients.

FIG. 13 is a diagram showing a configuration example of a code table A for counting effective coefficients.

FIG. 14 is a diagram illustrating another configuration example of a gradual area determination unit.

FIG. 15 is a flowchart illustrating an operation relating to a main part of the present invention of an image data encoding device according to another embodiment of the present invention.

FIG. 16 is a basic block diagram of a conventional image data encoding device.

[Explanation of symbols]

 1,11 Effective coefficient counting means 1a, 11a First counting means 1b, 11b Second counting means 2,13 Coding condition setting means 2a Detecting means 2b Judging means 2c setting means 3,14 Coding means 12 Block number counting means

Claims (5)

[Claims] 1. An original image is divided into a plurality of blocks composed of a plurality of pixels, the gradation values of a plurality of pixels in the plurality of blocks are orthogonally transformed for each block, and then the orthogonal transformation is performed. An image data encoding device that quantizes the obtained orthogonal transform coefficient, and encodes the quantized coefficient obtained by the quantization, wherein the number of appearances of the quantized effective coefficient is the plurality of blocks. Coefficient counting means (1) for counting, and coding condition setting means for setting a coding condition based on the number of appearances of the effective coefficients for each of the plurality of blocks counted by the effective coefficient counting means (1). (2) and an encoding means (3) for encoding the original image using the encoding conditions set by the encoding condition setting means (2)
An image data encoding device comprising: 2. The encoding condition setting means (2) includes a detection means (2a) for detecting a block in which the number of appearances of the effective coefficient is smaller than a predetermined value, and a detection result of the detection means (2a) in advance. A determination means (2b) for determining whether or not a predetermined condition is satisfied, and a setting means (2c) for setting the coding condition based on the determination result of the determination means (2b). The image data encoding device according to claim 1, characterized in that 3. The effective coefficient counting means (1) includes first counting means (1a) for counting the number of blocks in which the number of appearances of the effective coefficient is less than a predetermined value, and all the blocks of the original image. A second counting means (1b) for counting the number of appearances of the effective coefficient, and the encoding condition setting means (2) and the counting result of the first counting means (1a) and the second counting means. The image data coding apparatus according to claim 1, wherein the coding condition is set based on the counting result of the counting means (1b). 4. The original image is divided into a plurality of blocks composed of a plurality of pixels, the gradation values of a plurality of pixels in the plurality of blocks are orthogonally transformed for each block, and then the orthogonal transformation is performed. An image data encoding device that quantizes the obtained orthogonal transform coefficient and encodes the quantized coefficient obtained by the quantization, wherein the number of appearances of the quantized effective coefficient for each of the plurality of blocks. And an effective coefficient counting means (11) for counting, and a block number counting means (11) for counting the number of consecutive blocks in which the number of appearances of the effective coefficient is less than a predetermined value based on the counting result of the effective coefficient counting means (11). 12) and based on the counting result of the block number counting means (12),
A coding condition setting means (13) for setting a coding condition and a coding means (1) for coding the original image using the coding condition set by the coding condition setting means (13).
4) An image data encoding device comprising: 5. The effective counting means (11) includes a first counting means (11a) for counting the number of blocks in which the number of appearances of the effective coefficient is smaller than a predetermined value, and the effective number in the entire original image. A second counting means (11b) for counting the number of appearances of the coefficient, wherein the block number counting means (12) appears the effective coefficient based on the counting result of the first counting means (11a). The number of blocks whose number is smaller than a predetermined value is counted, and the coding condition setting means (13) and the counting result of the block number counting means (12) and the second counting means (11).
The image data encoding apparatus according to claim 4, wherein the encoding condition is set based on the counting result of b).