JPH05207290A - Picture data coder and coding method - Google Patents

Abstract

PURPOSE:To compress data efficiently with respect to the coder and its method applying 2-dimensional orthogonal transformation to picture data and coding the data. CONSTITUTION:The coder is provided with a quantization circuit generating section 1, a DC block detection section 3 detecting a DC block whose AC component quantization coefficient is all zero, a coding procedure decision section 4 calculating an evaluation figure based on a DC block number or its consecutive length and deciding the division coding of a variable length coding by a DC component only and a variable length coding of an AC component only as to one pattern based on the DC block number or the evaluation figure or consecutive variable length coding of the DC component and the AC component for each block or non-division coding, and a variable length coding section 5 applying variable length coding to a quantization coefficient according to the coding procedure decided by the coding procedure decision section 4.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a code for image data which is compressed by dividing a multi-valued image into blocks composed of a plurality of pixels, and two-dimensionally orthogonally transforming and coding the pixel data in the blocks. The present invention relates to an encoding device and an encoding method.

Since image data has a large amount of information, it is necessary to efficiently encode and compress the data in order to store it in a storage device or transfer it to another device at high speed. As a method of efficiently compressing image data, for example, there is an adaptive discrete cosine transform coding method.

Adaptive Discrete Cosine Transform Coding System (Ada
ptive Discrete Cosine Tra
The image is divided into blocks each having 8 × 8 pixels, and the pixel level of each block is a two-dimensional discrete cosine transform method (discrete cosine transform).
nsform (hereinafter abbreviated as DCT)), the coefficient is converted into a coefficient of spatial frequency distribution, quantized by a threshold adapted to the visual sense, and the obtained quantized coefficient is encoded by a Huffman table obtained statistically.

[0004]

2. Description of the Related Art FIG. 9 shows a conventional encoder. The figure shows the configuration of a conventional two-dimensional discrete cosine transform system.

In the figure, reference numeral 50 denotes a DCT transforming unit, which performs multidimensional image data two-dimensional discrete cosine transform and D
It outputs the CT coefficient. A quantizer 51 compares the DCT coefficient with a quantization threshold and quantizes the DCT coefficient. A quantization threshold table 52 stores the quantization threshold. Reference numeral 53 is a variable-length coding unit that performs variable-length coding on the quantized coefficient by referring to a Huffman code table (Huffman table). 54 is a Huffman code table (Huffman table).

The operation of the configuration of FIG. 9 will be described with reference to FIG. FIG. 10 is an explanatory diagram of a DCT coefficient encoding method. Figure (a) shows an example of quantized coefficients.

The figure shows the coefficients obtained by quantizing the DCT coefficient obtained by DCT transforming the image data of 1 block 8 × 8 pixels by comparing it with the quantization threshold value. In the figure, 60
Represents a DC component in the spatial frequency of one block of image data, and represents a horizontal frequency component (AC component) in the rightward direction and a vertical frequency component (AC component) in the downward direction.

FIG. 1B shows the scanning order of the quantized coefficient.
The figure shows the order of scanning a block of pixels to encode the quantized coefficients. The data to be encoded starts from the pixel of the DC component, scans in zigzag as shown in the figure, and sequentially encodes the quantized coefficients of the AC components in the high-order horizontal and vertical directions.

FIG. 1C shows an example of data to be encoded. The figure shows the coding target data obtained as a result of the zigzag scanning of the quantized coefficient in (a). As shown in the figure, starting from the DC component “5”, the AC components are “−2”, “−3”, “1” ...
Coding is performed in the order of and toward higher-order AC components. The array of "000" is run length data, and 0 is arrayed up to the last pixel of the scanning of one block after the ninth data (-1). Therefore, the terminal code "EOB" of one block is used.
Is added to mark one block of data as a delimiter.

The data to be encoded shown in the figure is sequentially encoded by referring to the Huffman table to obtain compressed data. FIG. 11 is a diagram showing an array of quantized coefficients and one screen block.

In the figure, reference numeral 70 is an example of a quantized coefficient, which is an example in which all AC components are 0 (hereinafter, a block in which all AC components are 0 is referred to as a DC block). 71
Shows an example of the arrangement of blocks on one screen, and shows the case where there are 5 blocks in the horizontal direction and a plurality of blocks in the vertical direction. It is assumed that block 1 to block 4, block 7 to block 10 are DC blocks, and block 5 and block 6 have non-zero quantized coefficients for AC components (hereinafter, non-zero quantized for AC components). Blocks with coefficients are called AC blocks).

In block 1, 72 represents the DC component of block 1. In the block 5, 73 is a two-dimensional AC component in the horizontal direction of the block 5, and 74 is A of the block 5.
It represents the highest harmonic component of the C component.

FIG. 12 shows a conventional encoding method. The figure shows a conventional encoding method in the case of encoding the quantized coefficients in the example of the block arrangement of one screen in FIG.

As shown in the figure, the data of block 1 to the data of the final block are sequentially encoded. Then, each block starts from the DC component and is A by zigzag scanning.
The C component is sequentially encoded. Then, in each block, up to the final data (the last non-zero data in the case of performing zigzag scanning of each block, in which 0s are lined up to the final pixel thereafter) is the encoding target, and the subsequent blocks Up to the last pixel (0 is lined up) is not coded, and "E
OB "is added to make a block delimiter.
Then, continuous data of 0s in the data portion to be encoded is run-length. If the last pixel of the block is not 0, the data of all the pixels are to be encoded, and E
No OB code is added.

In the figure, block 1 has a DC component of 5 and A
Since all C components are 0, 5 is arranged at the beginning and then EO
B is added to make one block of encoding target data. Similarly, in blocks 2 to 4, only the DC component is non-zero and the AC components are all 0, so that each block,
EOB is added after the DC component, and the data is coded for each block.

The block 5 includes a DC component "5" followed by A
Since the C component is "2", the AC component 2 is arranged following the first 5 and the AC component is sequentially encoded by zigzag scanning. Since the last non-zero data "1" is followed by 0, the last non-zero data "1" is the data to be encoded, and EOB is added to it as a block delimiter. The block 6 has a DC component and then an AC component 4, and since the last quantized coefficient of the block is "1", the AC component is sequentially encoded from "2" and up to the last "1". Since the block 6 knows the block delimiter without the end code EOB of the block, the head data of the next block following the data of the block 6 is set as the encoding target data without adding the EOB.

Similarly, in blocks 7 to 10, only the DC component is non-zero and the AC components are all 0. Therefore, following the DC component, EOB is sequentially encoded as encoding target data.

[0018]

In the conventional encoding method, only the DC component is non-zero in a portion where there is no image change, such as the background portion of image data of animation.
Thus, blocks in which the AC components are all 0 are continuous. And
Following the DC component encoding, the procedure of encoding "EOB" repeatedly appears, and "EOB" had to be variable-length encoded each time, which was inefficient.

An object of the present invention is to provide an image data encoding device and an encoding method capable of efficiently encoding when a block in which all AC components are 0.

[0020]

According to the present invention, an evaluation index based on the number of blocks of which all the AC components are 0 or the number of consecutive blocks in one screen block is calculated, and the number of blocks or evaluation index is compared with a threshold value. If it is larger than the threshold value, the coding procedure is divided into a procedure for continuously coding only the DC component of each block for one screen and a procedure for continuously coding only the AC component of each block. Encoding is performed (hereinafter, the DC component and the A
The method of dividing into C components and encoding is referred to as division encoding).

Prior to the basic configuration of the present invention of FIG. 1, FIG.
The encoding method of the present invention will be described below. 2 is shown in FIG.
The division encoding method in the encoding method of the present invention will be described with respect to the example (71) of the block arrangement of one screen.

In the figure, reference numeral 10 denotes a DC component, which indicates that only the DC component of each block is taken out in block order as data to be coded and sequentially coded. 11
Is an AC component, and indicates that only the AC component of each block is continuously subjected to zigsag scanning for each block in the block order and sequentially encoded. Then, when the AC component is the data to be encoded, the block in which the AC component is all 0 is E
An OB code is attached to make continuous EOBs into run lengths.

FIG. 1 shows the basic configuration of the present invention. In the figure, reference numeral 1 denotes a quantized coefficient creating unit, which DCT-transforms image data block by block, quantizes the image data by comparing it with a quantization threshold value, and creates a quantized coefficient for each block. Two
Is a coding means for coding the quantized coefficient created by the quantized coefficient creating section. A DC block detector 3 detects a DC block. Four
Is a coding procedure determining unit, which calculates an evaluation index based on the number of DC blocks or the number of consecutive DC blocks for one screen, compares the evaluation index with a threshold value, and performs division coding, or performs block coding as usual. It is determined for each time whether the DC component and the AC component are continuously encoded. Reference numeral 5 denotes a variable length coding unit, which performs variable length coding of the quantized coefficient according to the coding procedure determined by the coding procedure determination unit 4 with reference to the Huffman code table.

[0024]

The operation of the basic structure of the present invention will be described with reference to FIG. FIG. 3 shows a processing flow of the basic configuration of the present invention.

Description will be given according to the numbers shown. Please refer to FIG. (1) The quantized coefficient creating unit 1 quantizes a DCT coefficient created by DCT transforming image data for each block with reference to a quantization threshold value, and creates a quantized coefficient for each block.

(2) The DC block detector 3 inputs the quantized coefficient and determines whether or not the AC component has a non-zero value. (3) The coding procedure determination unit 4 determines whether there are many blocks in which the non-zero coefficient is composed only of DC (calculates an evaluation index based on the number of DC blocks or the number of consecutive DC blocks, and compares it with a threshold value. To). If the number of blocks or the evaluation index is larger than the threshold, proceed to (4), and if it is smaller, (8)
Proceed to.

(4), (5) The variable length coding unit 5 inputs the quantized coefficient and performs variable length coding on the DC component until the coding of the quantized coefficient of all blocks of the image is completed. (6), (7) The variable length coding unit 5 inputs the quantized coefficient and sequentially codes only the AC component for all blocks.

(8), (9) The variable length coding unit 5 continuously variable length codes the DC component and the AC component for each block for each block (hereinafter referred to as non-division coding).

[0029]

FIG. 4 shows an embodiment of the apparatus configuration of the present invention. In the figure, 15 is a DCT transform unit, 16 is a quantization unit, 17
Is a quantization threshold table, which stores the quantization threshold. Reference numeral 18 denotes an output direction selection unit which inputs the quantized coefficient from the quantization unit 16 and switches the output direction to the DC block detection unit 19 or the variable length coding unit 21. Reference numeral 19 is a DC block detector,
The DC block is detected by determining whether or not the AC component has a non-zero value. Then, the DC block detection unit 1
9 is a comparison means for comparing the AC component of the quantized coefficient with zero,
Counting means for counting the number of judged blocks and whether the judged block is a DC block, or A
It is provided with output means (not shown) for outputting an identification signal indicating whether the block is a C block. Reference numeral 20 denotes an encoding procedure determination unit that calculates an evaluation index based on the number of DC blocks detected by the DC block detection unit 19 or the number of consecutive DC blocks, compares the evaluation index with a threshold value, and divides and encodes the DC component and the AC component. It is determined whether to use the non-divisional encoding. Reference numeral 21 denotes a variable length coding unit, which inputs the quantized coefficient and performs variable length coding of the quantized coefficient according to the coding method determined by the coding procedure determination unit 20. Reference numeral 22 is a Huffman code table, and 23 is a timing control unit, which controls the timing of starting and ending the operation of each unit.

FIG. 5 shows a processing flow of the timing control section in the apparatus configuration of the present invention. The operation of the timing control unit will be described according to the processing flow of the figure (see FIG. 4).

(1) The timing control unit 23 instructs the output direction selection unit 18 so that the quantized coefficient output from the quantization unit 16 is input to the DC block side. (2) The timing control unit 23 uses the DCT for each block.
The changing unit 15 is instructed to perform DCT transform, and the quantizing unit 16 is instructed to perform quantization.

(3) It is judged whether or not the quantized coefficient of the block for one screen has been obtained. If the quantized coefficient for one screen is obtained, proceed to (4). If the quantized coefficient for one screen is not obtained, DCT transform and quantization of the next block are instructed (for the next block, see (2) Process).

(4) The timing controller 23 sets the output direction selector 18 (DMPX) so that the output of the quantizer 16 is input to the variable length encoder 21. (5) The number of times the variable-length coding unit 21 has instructed by the coding procedure determination unit 20 (twice for divided coding, once for non-divided coding), and for the DCT transform unit 15. The DCT transform for one screen, the quantization unit 16 is instructed to perform the quantization of the DCT transform, and the variable length coding unit 21 is instructed to perform the variable length coding.

As a result, in the case of division coding, DCT
The transforming unit 15 performs DCT transform for one screen twice, and the quantizing unit 16 compares the DCT coefficient of each block output from the DCT transforming unit 15 with the threshold value of the quantizing threshold value table 17 to quantize the coefficient. To create. Then, the output direction selection unit 18 (DMPX) inputs the quantized coefficient created by the quantization unit 16 to the variable length coding unit 21. Variable length coding unit 21
For variable quantized coefficients for the first input screen, only the DC component of each block is variable length coded. Further, regarding the quantized coefficient for one screen of the next time, only the AC component of each block is variable length coded.

In the case of non-divisional coding, the DCT transform unit 1
5 performs DCT conversion of image data for one screen only once, and inputs it to the quantization unit 16. The quantizer 16 uses the DCT
A quantized coefficient for one screen is created based on the DCT coefficient sent from the conversion unit 15. Further, the quantized coefficients for one screen are transferred to the variable length coding unit 21. Then, the variable length encoding unit 21 non-divisionally encodes the quantized coefficients for one screen for each block.

FIG. 6 is a processing flow of the DC block detector of the present invention. In the figure, 21 'is comparison means, 22'
Is a counting means, and 23 'is an output means. The processing flow will be described according to the numbers shown.

(1) According to the instruction of the timing control unit,
Read the quantized coefficient of AC component. (2) The comparison means 21 'compares the value of the quantized coefficient with 0 to determine whether the AC component is 0. If 0
Proceed to (3), and if not 0, proceed to (5).

(3) The counting means 22 'counts the determined number of pixels, and when the count reaches 64, one block has been completed, so the procedure proceeds to (4). If the count is less than 64, return to (2) and judge the next pixel.

(4) The output means 23 'has one block of A
Since the C component was all 0, the DC block detection signal D
Output C as "1". Then, a signal indicating the end of processing for one block is output to the timing control unit.

(5) The block of the output means 23 'is AC
Since it is a block (a block having a non-zero AC component), the DC block detection signal DC is output as "0". Then, a signal indicating the end of processing for one block is output to the timing control unit.

FIG. 7 is a diagram showing an embodiment of the coding procedure determining unit of the present invention. (a) is an embodiment of the encoding procedure determination unit (1)
FIG. In the figure, reference numeral 31 is a DC block number counting unit for counting the DC block detection signals (1 for a DC block, 0 for an AC block) detected by the DC block detection unit. Reference numeral 32 denotes a threshold value holding unit, which holds a threshold value for determining whether to perform divided coding or non-divided coding. Reference numeral 33 is a comparison unit, which compares the number of DC blocks with a threshold value. Reference numeral 34 denotes a coding division number determination unit, which, in the case of division coding, encodes the DC component and AC.
Outputs a coding division number 2 for instructing division encoding twice for component encoding, and in the case of non-division encoding, instructs DC component and AC component continuously for one encoding. The coding division number 1 is output.

FIG. 6B shows an embodiment (2) of the encoding procedure determining unit.
Is. In the figure, reference numeral 40 denotes a DC block continuous number counting unit, which counts the continuous number of DC blocks. Reference numeral 41 denotes an address generator, which is a weighting coefficient holding unit 4
An address designating a weighting coefficient of 2 is generated. A weight coefficient holding unit 42 holds a weight coefficient to be applied to the number of DC blocks according to the number of consecutive DC blocks. Reference numeral 43 denotes a data DC degree calculating unit, which calculates an evaluation index (DC degree) for division coding by applying a weighting factor to the number of consecutive DC blocks for each consecutive DC block. Reference numeral 44 is a threshold value holding unit, and 45 is a comparison unit for comparing the evaluation index calculated by the DC degree calculation unit 43 with the threshold value. Reference numeral 46 is a coding division number determination unit.

In the configuration shown in the figure, the DC block continuous number counting section 40 counts the DC block detection signal to count the continuous number of DC blocks. Then, the address generation unit 4 is operated so as to specify the address of the weight coefficient holding unit 42 that stores the weight count determined based on the number of consecutive times.
Instruct 1 to generate an address. The address generation unit 41 designates the designated address for the weight coefficient holding unit 42. The weighting factor holding unit 42 outputs the weighting factor of the designated address to the DC degree calculating unit 43.

The DC degree calculating unit 43 multiplies the continuous number counted by the DC block continuous number counting unit 40 by a weighting coefficient. Then, for one screen, the result of multiplying the number of consecutive times is cumulatively added to calculate the evaluation index (DC degree). Then, the comparison unit 45 compares the threshold value held by the threshold value holding unit 44 with D
The evaluation index calculated by the C degree calculation unit 43 is compared. If the evaluation index is larger than the threshold value, the coding division number determination unit 46 outputs the coding division number 2 to instruct division coding, and the evaluation index is the threshold value. If it is smaller, the encoding division number determination unit 46 outputs the encoding division number 1 to instruct non-division encoding.

The evaluation index is calculated by multiplying a predetermined integer N by a negative coefficient when the number of consecutive DC blocks is shorter than N, and when a DC block whose number of consecutive longer than N appears. Try to multiply by 0 or a positive coefficient. When the number of consecutive DC blocks is smaller than N, the coefficient is such that the negative coefficient is constant regardless of the length of the consecutive number, or the absolute value of the negative coefficient is increased as the number of consecutive is shorter. Alternatively, when the number of consecutive DC blocks is larger than N, the positive coefficient may be constant regardless of the length of the consecutive number, or the absolute value of the positive coefficient may be increased as the number of consecutive is longer. ..

FIG. 8 is a processing flow of the variable length coding unit of the present invention. In the figure, numeral 47 is a coding method instructing means in the variable length coding unit, which judges whether the number of coding divisions is two or one. Reference numeral 48 denotes an encoding means, which follows the instruction of the encoding method instruction means 47.
It is divided or non-divided.

The processing flow will be described according to the numbers shown. (1) The processing is started according to the instruction of the timing control unit, and the encoding method instruction unit 47 determines whether or not there is a division encoding instruction from the encoding procedure determination unit. If the number of encoding divisions is 2, it is an instruction of division encoding, so proceed to (2),
If the number of coding divisions is 1, it means non-divisional coding.
Proceed to (4).

(2) Since the instruction of the encoding method instructing means is divisional encoding, the encoding means 48 selects only the DC component of each block from the quantized coefficients input from the quantizer, and changes it. Long code.

(3) Variable length coding is performed by selecting only the AC component of each block from the quantized coefficients input from the quantizer. When the coding for one screen is completed, a signal indicating that is output to the timing control unit, and the process is completed.

(4) Since the encoding instruction of the encoding method instructing means 47 is non-divisional encoding, the encoding means 48 successively and continuously quantizes the quantized coefficients input from the quantizer. Turn into. When the coding for one screen is completed, a signal indicating that is output to the timing control unit, and the process is completed.

[0051]

According to the present invention, a block consisting of all AC components and only DC components is consecutive in a quantized coefficient obtained by two-dimensional orthogonal transformation, such as a character image or a background portion of an animation image. It is possible to efficiently code image data that often appears as a result.

[Brief description of drawings]

FIG. 1 is a diagram showing a basic configuration of the present invention.

FIG. 2 is an explanatory diagram of means for solving the problems of the present invention.

FIG. 3 is a diagram showing a processing flow of a basic configuration of the present invention.

FIG. 4 is a diagram showing a device configuration example of the present invention.

FIG. 5 is a diagram showing a processing flow of a timing control unit of the present invention.

FIG. 6 is a diagram showing a processing flow of a DC block detection unit of the present invention.

FIG. 7 is a diagram showing an embodiment of a coding procedure determination unit of the present invention.

FIG. 8 is a diagram showing a processing flow of a variable length coding unit of the present invention.

FIG. 9 is a diagram showing a conventional encoding device (two-dimensional discrete cosine transform method).

FIG. 10 is an explanatory diagram of a conventional quantization coefficient encoding method.

FIG. 11 is a diagram showing an array of quantized coefficients and blocks of one screen.

FIG. 12 is a diagram showing a conventional encoding method.

[Explanation of symbols]

 1: Quantization coefficient creation unit 2: Encoding means 3: DC block detection unit 4: Encoding procedure determination unit 5: Variable length encoding unit

Claims (6)

[Claims] 1. Input multi-valued image data, divide into a block composed of a plurality of pixels, perform a two-dimensional orthogonal transformation of the gradation value of the pixel for each divided block, and quantize the obtained coefficient. A quantized coefficient creation unit (1) and a DC block detection unit (3) that inputs a quantized coefficient and detects a DC block in which the quantized coefficients of AC components are all zero
And an evaluation index is calculated based on the number of DC blocks in one screen or the number of consecutive DC blocks, and variable length coding of only DC components and variable length coding of only AC components of all blocks of one screen are performed. An encoding procedure determination unit (4) for determining whether to perform division encoding for division and encoding, or non-division encoding for continuously variable length encoding DC components and AC components for each block; A coding device for image data, comprising: a variable length coding unit (5) for variable length coding the quantized coefficient according to the division method decided by the coding procedure decision unit (4). 2. Multi-valued image data is input, divided into blocks composed of a plurality of pixels, the gradation values of the pixels are two-dimensionally orthogonally transformed for each divided block, and the obtained coefficients are quantized. In the method for encoding image data, which comprises a quantized coefficient creating unit (1) and an encoding means (2) for encoding the quantized coefficient, the encoding means (2) inputs the quantized coefficient, A DC block in which the AC components of the quantized coefficients are all zero is detected to detect the number of the DC blocks in one screen, and an evaluation index is obtained based on the number of the DC blocks or the number of consecutive blocks, and the number of the blocks is calculated. Alternatively, the evaluation index is compared with a threshold value, and if it is larger than the threshold value, it is divided into coding of only DC components and coding of only AC components for all blocks of one screen, and the number of blocks or the evaluation index is greater than the threshold value. If small, each block Method of encoding image data, characterized by continuously encoded DC and AC components for each. 3. The evaluation index according to claim 1, wherein, when the number of consecutive blocks of which AC components are all zero is larger than a predetermined integer N, the weight is applied to the consecutive numbers as zero or a large positive weight. When the number of consecutive blocks is smaller than N, it is obtained by acting on a negative or small weight consecutive number. 4. The weight according to claim 3, wherein the shorter the number of consecutive blocks, the smaller the weight, or the larger the absolute value of the negative weight, the larger the number of consecutive blocks, the larger the weight. Image data encoding device. 5. The evaluation index according to claim 2, wherein when the number of consecutive blocks of which the AC components are all zero is larger than a predetermined integer N, the weight is applied to the consecutive numbers as zero or a large positive weight. When the number of consecutive blocks is smaller than N, a negative or small weight is applied to the number of consecutive blocks to obtain the image data encoding method. 6. The weight according to claim 5, wherein the shorter the number of consecutive blocks is, the smaller the weight is, or the negative value is, the absolute value thereof is increased, and the longer the number of consecutive blocks is, the larger the weight is. Image data encoding method.