JPH0670175A - Method and device for encoding picture data - Google Patents

Abstract

PURPOSE:To obtain the code data of small data quantity while suppressing the deterioration of picture quality in respect of a method and a device for encoding picture data to compress and encode the picture data. CONSTITUTION:The picture data is divided into one or plural pieces of blocks in a step S1, and an orthogonal transformation coefficient is calculated by giving orthogonal transformation to the divided data in the step S2, and a quantized orthogonal transformation coefficient is calculated by quantizing the orthogonal transformation coefficient in the step S3. In the step S4, the absolute value of the quantized orthogonal transformation coefficient Sq is compared with a predetermined threshold Th, and a non-zero coefficient below the threshold is selected, and in the step S5, it is changed into a zero coefficient in the way of S - or + = 0, and finally in the step S6, the changed quantized orthogonal transformation coefficient is entropy-encoded.

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 method and apparatus for compressing and coding image data. recent years,
In order to efficiently handle image data having a very large amount of data, a coding technique for compressing image data has begun to be used. Although there are many methods for encoding image data, there is a special article in the magazine interface, December 1991, "Understanding and Application of Image Data Compression (pp. 129-23).
As described in "Page 1)", it is obtained by a process of dividing the image data into blocks, an orthogonal transformation process of performing an orthogonal transformation on the block-divided images, and an orthogonal transformation process in view of high compression efficiency and the like. The JPEG method, which is a combination of a quantization process that performs a quantization process on the obtained orthogonal transform coefficient and an entropy coding process that performs entropy coding on the quantized transform coefficient obtained by the quantization process, is used internationally. It's being done.

Even with such an encoding method, image data can be compressed into a small amount of data. However, the amount of data obtained by encoding (hereinafter referred to as “coded data”) is still much larger than that of data such as documents, and development of a more efficient encoding method is required. ing. Further, when the coded data is transmitted to a remote place via a communication line, it may be necessary to encode the coded data into a smaller amount of data for transmission due to reasons such as reduction of transmission time and usage fee of the communication line. Therefore, there is a demand for the development of a coding method for efficiently storing and transmitting code data by performing coding so that the data amount can be significantly reduced even if the image quality is slightly degraded, if necessary. .

[0003]

2. Description of the Related Art FIG. 15 is a flowchart showing a conventional image data encoding method. In FIG. 15, first, in step S1, the image data is divided into one or a plurality of blocks, and in step S2, the image obtained by the block division is subjected to orthogonal transformation to calculate orthogonal transformation coefficients. In step S3, the quantized orthogonal transform coefficient is calculated by quantizing the orthogonal transform coefficient obtained by the orthogonal transform. Finally, in step S4, entropy coding is performed on the quantized orthogonal transform coefficient to obtain coded data.

In such an encoding method, as a method of increasing the compression rate, there is a method of increasing the quantization threshold value. The reason why the compression rate can be increased by increasing the quantization threshold is as follows. Entropy coding has a characteristic that the smaller the data variation, the more efficiently the data can be compressed. As is widely known, the values of the orthogonal transform coefficient are concentrated on zero and small numerical values. Naturally, the quantized orthogonal transform coefficient quantized by dividing the orthogonal transform coefficient by the quantization threshold also concentrates on the numerical values of zero and the absolute value.

When the quantization threshold value is increased, the degree of concentration of the quantized orthogonal transform coefficient on the values of zero and small absolute value increases. As described above, in entropy coding, the smaller the variation of data, the more efficient the compression is.Therefore, the compression rate can be increased by increasing the quantization threshold, and as a result, the amount of data after entropy coding can be increased. Can be reduced.

[0006]

However, in the conventional image data coding method in which the quantization threshold is increased to reduce the data amount, when the quantization is performed with a large quantization threshold, the quantized orthogonal transform coefficient Since the number of steps for expressing is reduced, the image quality of the image restored from the coded data becomes extremely low.

An object of the present invention is to suppress deterioration of image quality,
An object of the present invention is to provide an image data encoding method and apparatus capable of obtaining encoded data with a small amount of data.

[0008]

1 and 2 are explanatory views of the principle of the present invention. [First Invention] A first invention of the present application is an image data encoding method for encoding image data, wherein an orthogonal transform coefficient is applied to block data obtained by dividing image data into one or a plurality of blocks. Of the quantized orthogonal transformation coefficient obtained in the quantization process and the quantization process of calculating the quantized orthogonal transformation coefficient by quantizing the orthogonal transformation coefficient obtained in the orthogonal transformation process. The absolute value is compared with a predetermined threshold value, the comparison process of selecting a non-zero coefficient equal to or less than the threshold value, the coefficient value changing process of setting the non-zero coefficient selected in the comparison process to zero, and the coefficient value changing process And an entropy coding process of performing entropy coding on the quantized orthogonal transform coefficient which has been completed.

Here, it is desirable that the threshold used in the quantization process is a threshold corresponding to the human visibility for each orthogonal transform coefficient, and the threshold used in the comparison process is the human visibility for each quantized orthogonal transform coefficient. It is desirable to set the threshold according to [Second Invention] A second invention of the present application is an image data encoding method for encoding image data, in which orthogonal transformation is performed by applying orthogonal transformation to block data obtained by dividing image data into one or a plurality of blocks. Select in the comparison process, the orthogonal transformation process of calculating the transformation coefficient, the comparison process of comparing the absolute value of the orthogonal transformation coefficient obtained in the orthogonal transformation process with a predetermined threshold value, and selecting the non-zero coefficient below the threshold value. Coefficient value setting process for setting the quantized orthogonal transform coefficient corresponding to the selected non-zero coefficient to zero, and the quantization process for calculating the quantized orthogonal transform coefficient by quantizing the orthogonal transform coefficient not selected in the comparison process. And an entropy coding process for performing entropy coding on the quantized orthogonal transform coefficients obtained in the coefficient value setting process and the quantization process.

Also in this case, the threshold used in the comparison process is
It is desirable to use a threshold value corresponding to the human visibility for each orthogonal transformation coefficient.

[0011]

First, in the image data coding method according to the first invention of FIG. 1, after obtaining the quantized orthogonal transform coefficient Sq in step S3, the quantized orthogonal transform coefficient is compared with a predetermined threshold Th in step S4. When Sq is less than or equal to the threshold Th, a process of replacing the quantized orthogonal transform coefficient value with Sq = 0 is newly added in step S5.

Generally, the quantized orthogonal transform coefficients are concentrated on the values of zero and small absolute values. It is known that in an actual image, the quantized orthogonal transform coefficient is very often zero. Therefore, the amount of data after entropy coding largely depends on the frequency of appearance of zero in the quantized orthogonal transform coefficient. In the present invention, the amount of data after entropy coding, which largely depends on the appearance frequency of zero of the quantized orthogonal transform coefficient, is selected by the operation of selecting only the quantized orthogonal transform coefficient having a smaller absolute value after quantization and replacing it with zero. Reduce.

In this case, since the frequency component corresponding to the quantized orthogonal transform coefficient with a small absolute value, which is replaced with zero, is lost, the image quality is slightly deteriorated when restored. However, since the accuracy (number of gradations) of the other non-zero coefficients that remain unselected does not change, the effect on the image quality is only a small effect due to the non-zero coefficient replaced by the zero coefficient with a small absolute value. It is possible to efficiently obtain coded data having a small amount of data while suppressing the decrease of the number of bits.

In the second invention of FIG. 2, the orthogonal transform coefficient S obtained by the orthogonal transform in step S2 is compared with a predetermined threshold Th in step S3. The step quantization process in S of step S5 is skipped, and the quantized orthogonal transform coefficient Sq is uniquely set to Sq = 0 in step S4. Therefore, similar to the first aspect of the invention, in addition to the significant reduction of the code data in which the deterioration of the image quality is minimized, the number of orthogonal transform coefficients to be subjected to the quantization processing can be significantly reduced, and the processing time is required in the quantization process. In the case of the system configuration, for example, when the quantization processing is performed by software, the processing time can be shortened significantly.

Of course, even if the quantization process is executed without skipping step S5 after setting the orthogonal transformation coefficient Sq to Sq = 0 in step S4, the same coding effect can be obtained.

[0016]

FIG. 3 is a block diagram of an embodiment showing an embodiment of the first invention of the present application. In FIG. 3, the image data encoding device according to the first aspect of the present invention includes a block division unit 10, an orthogonal transformation unit 12, a quantization unit 14, a quantization threshold table 16, and a coefficient value comparison unit 1.
8, a coefficient value changing unit 20 and an entropy coding unit 22.

The block division unit 10 divides the image data into blocks of, for example, 8 × 8 pixels shown in FIG. 5, and outputs the image data to the orthogonal transformation unit 12 in block units. The orthogonal transform unit 12 orthogonally transforms block image data by, for example, discrete cosine transform (hereinafter referred to as “DCT”) as orthogonal transform, and transforms the block image data shown in FIG. 5 to the orthogonal transform coefficient S of the spatial frequency distribution shown in FIG. Ie DC
The T coefficient S is converted and output to the quantization unit 14.

The quantizing unit 14 divides the input DCT coefficient S by a quantizing threshold value table 16 composed of threshold values shown in FIG. 7 determined by visual experiment to obtain a quantized orthogonal transform coefficient Sq, that is, a quantized DCT. A quantization operation for obtaining the coefficient Sq is performed. FIG. 8 shows a quantized DCT coefficient calculated by using the threshold of the quantization threshold table 16 shown in FIG. 7 for the DCT coefficient S shown in FIG.

The quantized DCT coefficients shown in FIG. 8 are all quantized DCT coefficients calculated from the DCT coefficients smaller than the quantization threshold, and the quantized DCT coefficients have values only in the DC component and a few AC components, that is, non-zero. It is generated as a coefficient. Quantized DCT obtained by the quantizer 14 as shown in FIG.
The coefficient Sq is sequentially given to the coefficient value comparison unit 18 and quantized D
The absolute value of the CT coefficient Sq is compared with a predetermined threshold Th. For the quantized DCT coefficient Sq whose absolute value is smaller than the threshold Th by this comparison, the coefficient value changing unit 20 changes the quantized DCT coefficient Sq smaller than the threshold Th to a zero coefficient.

That is, the quantized DCT coefficient Sq 'as the zero coefficient after the change is output with Sq' = 0. Of course, if the absolute value of the quantized DCT coefficient Sq is greater than or equal to the threshold Th, the quantized DCT coefficient Sq is output to the entropy coding unit 22 as it is. The coefficient value changing unit 20 based on the comparison result of the coefficient value comparing unit 18
As a result of the changing process by, the DCT coefficient shown in FIG. 8 is changed as shown in FIG. 9, for example. Quantized DC after modification of FIG.
In order to simplify the description, for example, the T coefficient has a threshold value Th that is compared by the coefficient value comparison unit 18 as, for example, Th = 10, and the absolute value is smaller than the threshold value Th = 10.
As shown in FIG. 9, all four coefficient values of −1 and −8 are changed to 0, 64 as a DC component and − as an AC component.
Only 16 and -27 remain, and all others have zero coefficient.

The entropy coding unit 22 performs entropy coding after changing the quantized DCT coefficient shown in FIG. 9 into one-dimensional data according to an operation order called zigzag scan shown in FIG. As this entropy coding method, for example, Huffman coding using a fixed table is performed. For example, the difference between the DC component 64 at the head of the quantized DCT coefficient after the change in FIG. 9 and the DC component of the immediately preceding block that has already been processed is subjected to variable length coding. For the AC component, a value of non-zero coefficient that is not zero is used as an index, and the number of zero coefficients up to the index is used as a stage to generate data that is a combination of the index and the stage and is variable length coded for each block.

The variable-length coded DC component and AC component are output as coded data by Huffman coding using a code table composed of a Huffman table created on the basis of statistics for each image. To do. FIG. 4 is a flow chart showing the processing operation of the embodiment of FIG. 3, in which the image is divided into blocks in step S1 and the block image divided in step S2 is subjected to orthogonal transformation, that is, DC.
T-transform and perform DC using the quantization threshold in step S3.
The quantized DCT coefficient is calculated by quantizing the T coefficient.

Then, in step S4, the quantized DCT coefficients Sq for one block are sequentially read out and compared with the threshold Th.
If it is smaller than the threshold Th, this quantized DC is calculated in step S5.
The T coefficient Sq is changed to Sq = 0 and a zero coefficient. Then, in step S6, the quantized DCT coefficient obtained as it is or the changed quantized DCT coefficient is entropy-coded,
After checking the end of one block in step S7, the next quantized DCT coefficient is selected in step S8, and similar processing is repeated.

When it is determined in step S7 that one block has been processed, the process proceeds to step S9, in which a block end identifier is generated and entropy-encoded. Then, in step S10, it is checked whether all blocks have been processed. Block is selected and the process from step S4 is repeated. When it is determined in step S10 that all blocks have been processed, a series of encoding processes ends.

In the coding process of FIG. 4, block quantization, DCT conversion and quantization are performed on all blocks of one screen in steps S1 to S3, and then the quantized DCT coefficient is changed in block units. But step S1
The processing from 1 may be repeated for each block. As described above, in the first invention of the present application, a threshold value is further added to the quantized DCT coefficient as shown in FIG. 8, which is obtained by the quantization process for calculating the quantized DCT coefficient using the quantization threshold value. Since the non-zero coefficient whose absolute value is smaller than the threshold value Th is changed to zero coefficient in comparison with Th, the changed quantized DCT coefficient is as shown in FIG.
The degree of concentration of the coefficients to zero and small numerical values increases, and the variation in data in entropy coding decreases, so that the compression rate can be improved and the data amount of finally obtained code data can be reduced.

In this case, the quantized DCT coefficient shown in FIG.
Even if the value is changed by comparison with the threshold value Th, the value of the quantized DCT coefficient that has not been changed remains unchanged.
Even if the amount of data is reduced, the influence of deterioration on the image quality can be suppressed to the necessary minimum. In this embodiment, a value corresponding to human visibility is used as the quantization threshold. Therefore, the luminosity per level of the quantized DCT coefficient can be treated as a constant value, and the threshold value Th can be a constant value. However, when a quantization threshold value that does not consider human visibility is used, it is desirable to use a threshold value Th that is weighted according to the visibility of the quantized DCT coefficient.

FIG. 11 is a block diagram of an embodiment of the second idea of the present application, in which the encoding device of this embodiment is a block division unit 1.
0, the orthogonal transformation unit 12, the coefficient value comparison unit 24, the coefficient value setting unit 26, the quantization unit 14, the quantization threshold table 16 and the entropy coding unit 22. The feature of this embodiment is that the coefficient value comparison unit 24 provided subsequent to the orthogonal transformation unit 12 compares the absolute value of the DCT coefficient S, which is the orthogonal transformation coefficient, with a predetermined threshold Th, and when it is smaller than the threshold Th, the coefficient value is smaller. The setting unit 26 does not perform the quantization process in the quantization unit 14, and directly sets the quantized DCT coefficient (quantized orthogonal transform coefficient) Sq = 0 to the entropy coding unit 22. .

Therefore, in the quantizing unit 14, the coefficient value comparing unit 24 has an absolute value of the DCT coefficient S which is equal to or larger than the threshold Th.
Only the quantization threshold value Q in the quantization threshold value table 16 may be divided to obtain the quantized DCT coefficient Sq.
4 can greatly reduce the number of DCT coefficients to be processed,
Efficient quantization processing can be realized when the quantization processing takes time.

FIG. 12 is a flow chart showing the processing operation of the embodiment shown in FIG. In FIG. 12, first, in step S1, the image data is divided into blocks, and in step S2, the block image is orthogonally transformed. For example, D by DCT
Convert to CT coefficient S. Then, in step S3, the absolute value of the DCT coefficient S is compared with the threshold Th. If it is smaller than the threshold Th, the quantized DCT coefficient Sq calculated from this DCT coefficient S is set to Sq = 0 in step S4, and the quantum of step S5 is set. The entropy coding is performed in step S6 without performing the coding process.

Of course, if the absolute value of the DCT coefficient S is equal to or more than the threshold Th in step S3, quantization processing is performed as usual in step S5 to obtain a quantized DCT coefficient, and entropy coding is performed in step S6. . Step S3 ~
The process of S6 is repeated until the end of one block is determined in step S7. When one block is completed, a block end identifier is generated and entropy-encoded in step S9, and the above process is completed in step S10. Repeat until you determine.

In the embodiment of FIG. 11, when the absolute value of the DCT coefficient S is smaller than the threshold value Th in the coefficient value comparing section 24, the quantization processing in the quantizing section 14 is skipped and forced. Although the coefficient value setting unit 26 sets the DCT coefficient Sq = 0, when the coefficient value comparing unit 24 selects a DCT coefficient whose absolute value is smaller than the threshold value Th, this DCT coefficient Sq
The same result can be obtained by changing S to S = 0 and then performing quantization in the quantization unit 14.

FIG. 13 shows one block of the DCT coefficient S obtained by the orthogonal transform unit 12 of FIG. 11, and in the next coefficient value comparison unit 24, for example, Th = Th is set as the threshold Th.
If 60 is set, it is determined that the DCT coefficients S other than those surrounded by the circles shown in FIG. 13 are smaller than the threshold Th, and the coefficient value setting unit 26 sets the corresponding quantized DCT coefficient Sq.
By receiving the setting of = 0, the DCT coefficient shown in FIG. 14 can be obtained.

Even for the quantized DCT coefficient shown in FIG. 14, the quantized DCT coefficient corresponding to the DCT coefficient having a small absolute value.
The coefficient is selectively changed to the zero coefficient, and the quantization D
By concentrating the CT coefficients on zero or a small absolute value, it is possible to reduce the amount of coded data by Huffman coding. Further, for the quantized DCT coefficients corresponding to the DCT coefficients equal to or more than the threshold Th, for example, the three DCT coefficients 368, −65, and −111 in FIG. Since it is required, the influence on the deterioration of the image quality is suppressed to the minimum, and the code data having a small data amount can be efficiently obtained.

Here, in the quantizing unit 14 provided in each of the embodiments of FIGS. 3 and 11, the quantizing threshold corresponding to the human visual sensitivity shown in FIG. 7 is used as the quantizing threshold table 16. is doing. That is, the quantized DCT coefficient of FIG. 8 is calculated using a large quantization threshold value for a quantized DCT coefficient of low human visibility, while a small quantization value is calculated for a quantized DCT coefficient of high human visibility. The quantized DCT coefficient is calculated as a component having the same importance with respect to human visibility even when the absolute value of the quantized DCT coefficient is constant. it can.

Further, in the embodiment of FIG. 11, the case where a constant threshold value is used as the threshold value Th used in the coefficient value comparison unit 24 is taken as an example, but this threshold value Th is also shown in FIG. Similar to the quantization threshold used in the quantization processing shown, it is desirable to use a threshold having a weight according to the human visibility. Specifically, the coefficient value comparison unit 2 in FIG.
In the case of 4, by using a threshold value corresponding to the human sensitivity per level of the DCT coefficient, that is, a large threshold value is used for the DCT coefficient with low human visibility, and a small threshold value is used for the DCT coefficient with high human visibility. , It is possible to minimize the influence of the image quality on the restored image.

[0036]

As described above, according to the present invention, it is possible to realize an encoding process capable of obtaining coded data with a small amount of data without increasing image quality deterioration of an image restored from coded data. Therefore, it is possible to appropriately cope with the shortage of the capacity of the code data storage unit and the time-constrained data transmission.

[Brief description of drawings]

FIG. 1 is an explanatory view of the principle of the first invention.

FIG. 2 is an explanatory view of the principle of the second invention.

FIG. 3 is a configuration diagram of an embodiment of the first invention.

FIG. 4 is a flowchart showing the processing operation of FIG.

FIG. 5 is an explanatory diagram of original image data for one block.

FIG. 6 is a DCT when the image signal of FIG. 5 is DCT-transformed.
Explanatory diagram showing coefficients

FIG. 7 is an explanatory diagram of a quantization threshold adapted to human vision.

8 is an explanatory diagram of a quantized DCT coefficient when the DCT coefficient of FIG. 6 is quantized using the quantization threshold value of FIG.

9 is an explanatory diagram of a quantized DCT coefficient in which a value smaller than a threshold Th among the quantized DCT coefficients in FIG. 8 is changed to a zero coefficient.

FIG. 10 is an explanatory view showing an operation sequence when converting the quantized DCT coefficient of FIG. 9 into one-dimensional data.

FIG. 11 is a configuration diagram of an embodiment of the second invention.

12 is an explanatory diagram showing the processing operation of FIG.

13 is DC compared with a threshold Th in the embodiment of FIG.
Explanatory drawing showing T coefficient

14 is an explanatory diagram of quantized DCT coefficients obtained according to the selection result of DCT coefficients in FIG.

FIG. 15 is a flowchart showing a conventional encoding method.

[Explanation of symbols]

 10: Block division unit 12: Orthogonal transformation unit 14: Quantization unit 18, 24: Coefficient value comparison unit 20: Coefficient value change unit 22: Entropy coding unit 26: Coefficient value setting unit

Claims (7)

[Claims] 1. An image data encoding method for encoding image data, wherein an orthogonal transform coefficient is calculated by applying an orthogonal transform to block data obtained by dividing the image data into one or a plurality of blocks. A process, a quantization process for calculating a quantized orthogonal transform coefficient by quantizing the orthogonal transform coefficient obtained in the orthogonal transform process, and an absolute value of the quantized orthogonal transform coefficient obtained in the quantization process. A comparison process of comparing with a predetermined threshold value and selecting a non-zero coefficient equal to or less than the threshold value, a coefficient value changing process of setting the non-zero coefficient selected in the comparison process to zero, and a change of the coefficient value changing process are performed. An entropy coding process for performing entropy coding on the completed quantized orthogonal transform coefficient, and an image data coding method. 2. The image data encoding method according to claim 1, wherein the quantization threshold value used in the quantization step is a quantization threshold value corresponding to human visibility for each orthogonal transform coefficient. Image data encoding method. 3. The image data encoding method according to claim 1, wherein the threshold value used in the comparison step is a threshold value corresponding to human visibility with respect to each quantized orthogonal transform coefficient. Data encoding method. 4. An image data encoding method for encoding image data, wherein an orthogonal transform coefficient is calculated by applying an orthogonal transform to block data obtained by dividing the image data into one or a plurality of blocks. Process, a comparison process of comparing the absolute value of the orthogonal transformation coefficient obtained in the orthogonal transformation process with a predetermined threshold value, and selecting a nonzero coefficient equal to or less than the threshold value, and a nonzero coefficient selected in the comparison process. A coefficient value setting step of setting the quantized orthogonal transform coefficient to zero, a quantization step of calculating a quantized orthogonal transform coefficient by quantizing an orthogonal transform coefficient not selected in the comparison step, An image data code having an entropy coding process for performing entropy coding on the quantized orthogonal transform coefficients obtained in the numerical setting process and the quantization process. Method. 5. The image data coding method according to claim 4, wherein the threshold value used in the comparing step is a threshold value corresponding to human visibility for each orthogonal transform coefficient. Method. 6. An image data coding apparatus for coding image data, wherein an orthogonal transform coefficient is calculated by applying an orthogonal transform to block data obtained by dividing the image data into one or a plurality of blocks. Means, a quantizing means for calculating a quantized orthogonal transform coefficient by quantizing the orthogonal transform coefficient obtained by the orthogonal transform means, and an absolute value of the quantized orthogonal transform coefficient obtained by the quantizing means. Comparison means for comparing with a predetermined threshold value and selecting a non-zero coefficient equal to or less than the threshold value, coefficient value changing means for setting the non-zero coefficient selected by the comparing means to zero, and change by the coefficient value changing means An image data encoding device, comprising: entropy encoding means for performing entropy encoding on the completed quantized orthogonal transform coefficient. 7. An image data coding apparatus for coding image data, wherein an orthogonal transform coefficient is calculated by applying an orthogonal transform to block data obtained by dividing the image data into one or a plurality of blocks. Means, comparing means for comparing the absolute value of the orthogonal transformation coefficient obtained by the orthogonal transformation means with a predetermined threshold value, and selecting a non-zero coefficient equal to or less than the threshold value; and a non-zero coefficient selected by the comparison means. Coefficient value setting means for setting the quantized orthogonal transform coefficient to zero, and quantizing means for calculating the quantized orthogonal transform coefficient by quantizing the orthogonal transform coefficient not selected by the comparing means, An image data code, comprising: entropy coding means for performing entropy coding on the quantized orthogonal transform coefficients obtained by the numerical value setting means and the quantizing means. Apparatus.