JPH03292081A - Coding and decoding device - Google Patents

Abstract

PURPOSE:To make the entire processing efficient by not scanning a coefficient data at the coding and not reproducing the data at the decoding side but terminating processing when a consecutive zero coefficient data at the end of zigzag scanning is in existence in a block to be coded or decoded. CONSTITUTION:For example, a coder 22 generates a Huffman code data from a data from a buffer 7 by using combination between a run length value of consecutive zero data and its succeeding non-zero data and outputs the result as a coding data 23. Then a count 18 of a counter 14 and a count 16 of a count register 6 are compared, when they are coincident, a count coincidence signal 17 is informed to the coder 22. Thus, in the code processing after one- dimensional array conversion, when the final data in a block is zero, the coding processing is attained without scanning the data. Thus, the processing efficiency of final consecutive zero data in zigzag scanning in a block is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a compression device for natural images including color images. Although the mainstream is the discrete cosine transform (Discrete Co51ne Tran
sform: abbreviated as DCT, hereinafter referred to as DCT)
This method is common, but this method uses NxM image data.
(N and M are the same value, 8 or 16 are common) Blocks are divided by LDCT, then quantized, the quantized data is converted into a one-dimensional array, and then encoded. Fig. 3 shows a method of scanning pixel data in a divided block. The basic configuration block diagram of the encoding process is shown in the figure.The basic configuration of the encoding process will be explained using FIG.
DCT rope that performs CT processing 63 is DCT coefficient database 6
4 is quantization processing performed on the coefficients after DCT. 65 is coefficient data 9 after quantization. 〇〇 is Huffman encoding. 67 is code data 68 is Huffman decoding. 69 is quantization that performs inverse quantization processing. -1 thought 70 is a DCT that performs inverse DCT processing
The image data 61 is divided into blocks, and the block data is subjected to DCT processing by a DCT unit 62 and converted to coefficient data 63.The coefficient data quantization unit 63 performs quantization processing.After quantization, the image data 61 is divided into blocks. The coefficient data becomes 65, which is encoded by the Huffman coding section 66 to become coded data 67. In the Huffman coding section 66, the coefficient data after quantization;
) Perform a zigzag scan as shown in ~\ Of the coefficients after DCT processing, the DC component (coefficient value at the top left of the block) is Huffman-encoded, and the DC component is removed <the AC component coefficient (other Huffman encoding is performed on the run length value, which is the number of consecutive data whose coefficient is zero after quantization (coefficient data), and the non-zero data value.And if the end of the block ends with a run of zero. is E○B (End Of Block) without encoding the run length value of the example.
) codes and completes the encoding process for that block.

In addition, during decoding processing (the coded data 67 is decoded into quantized coefficient data in a Huffman decoding section 68, and then inverse quantization processing is performed in a quantization-1 section 69). The DCT-1 unit 70 performs inverse DCT processing on the data to produce surface image data.In the process of the Huffman decoding unit 68, the coefficient data is decoded along the zigzag scan and the hE○B code is detected. In addition to decoding up to that point, it is also possible to decode up to the end of the block. I Niss O/I E C C C D C/
Niss C 2/Ta 97# Squirrel C 0-8°C C I T Niss C 9-■ ("JP
EG Technical 5specificat
ion, Revision 5” ISO/I
ECTTCI/SC2/WG8. CCITT SG
(2) An example of actual encoded data is shown in FIG. In this diagram
In the figure, even if only the encoding process of the AC component is shown, 1)
indicates the scanning of data within the block.The numbers outside the block in the figure indicate the X and Y coordinate values.More 2) indicates that the encoding process is performed by zigzag scanning within the blower. Indicates the code data for the minute, C(X, Y)
is the data that encodes the combination of the number of zero runs and the coefficient value that follows, X is the number of zero runs, and Y represents the quantized coefficient data value. f, DCT processing is usually performed using matrix calculations, so as shown in Figures 3 (2) and 3 (3), there are many cases where processing is performed by scanning in the horizontal or vertical direction. A scan similar to that performed on coefficient data is often used (~ In addition, zigzag scanning during Huffman encoding processing, as shown in Figure 13 (1), is performed before Huffman encoding processing or during decoding. It is necessary to temporarily store the processed data in a buffer. An example of its configuration is shown in Figure 5. In Figure 5, 51 is the coefficient data block after quantization.
2 is a transfer control signal Ji! indicating the transfer timing of the coefficient data, etc. P% 53 is an address generator that generates an address to be given to the buffer. 54 is a buffer that temporarily stores coefficient data. 55 is an address given to the buffer 54. Address 56 is a Huffman code. Decoder 57 is an atrea given to the buffer. 58 is a buffer. The coefficient data 59 that goes back and forth between the encoder and decoder is code data.The encoding process in this block diagram will be explained below.Coefficient data 51 after quantization
is once input to the buffer 54 for scan conversion.At that time, the address generator 53 generates an address to be stored in the buffer 54 from the data input timing obtained by the transfer control signal 52, and gives it to the buffer 54.At that time, the address is Even if the data is generated in the scanning order as shown in FIG. 3 (2) or FIG. tN, one-dimensional converted data 58 is obtained from the buffer 54 and encoded to obtain coded data 59. In this encoding process, the zigzag scan in prop 2 is performed at the end of the block. It must be done up to the coefficient of 8.
However, if the block ends with zero, that continuous zero data is not converted to code data, but is composed of code data up to the coefficient data before the last continuous zero data and E○B code. It is possible to generate encoded data without scanning the last consecutive data, but with this conventional method, it is necessary to check whether the zeros continue to the end. chicken
The field color of the encoding method using Mat-, DCT The second half of the zigzag scan has a high probability of becoming zero. It ends with a run of zero. In other words, the last zero run is not encoded and the fS EOB code is connected. In many cases, the decoding process ends in the following way (Fig. The encoded data 59 is input to the encoder/decoder 56, and the decoded data 58 is inputted to the buffer 54, and the address 57 at that time is output from the encoder/decoder 56 to perform the zigzag scan shown in FIG. 3(1). If the coded data to be decoded ends with an EOB code, it is necessary to set the data in the block following the reproduced data before EOB detection to zero until the end. This is because there is a possibility that data from the previous block remains.In this kind of decoding process, the L code processing machine also scans pixels that are not reflected in the actual code data, which is an inefficient problem solved by the invention. In the conventional example described above, there is a problem that the efficiency of the encoding/decoding process is not sufficient when there is continuous zero coefficient data at the end of the block zigzag scan to be encoded or decoded. The present invention takes account of this point and aims to provide an encoding device and a decoding device with high processing efficiency for such data. Image data is N×M (N,
Divide into blocks (M is a natural number), perform orthogonal transformation and quantization processing on the blocks, and convert the data into 1
A device that converts data into a dimensional array and performs encoding processing includes a pixel data processing means for processing input pixel data, a means for temporarily storing pixel data to be encoded, and a means for temporarily storing pixel data to be encoded. means for counting at least one or more non-zero pixel data to be encoded, means for temporarily storing the counted value, and storing the pixel data to be encoded from the temporary storage means. It has a means for reading into a dimensional array, a means for comparing count values, and a means for encoding the pixel data to be encoded. The pixel data processing means obtains the pixel data to be encoded and information that the pixel data is non-zero, counts it, temporarily stores the counted value, and stores the temporarily stored pixel data to be encoded as 1. When converting to a dimensional array and reading it out, the number of non-zeros in the data is counted again, and when the temporarily stored non-zero count value is reached, the encoding is terminated. converter Cyo converts image data to N0M (
(N, M are natural numbers) blocks, orthogonal transformation and quantization are performed on the blocks, the data is converted to one-dimensional array data, and an encoding device converts the encoded data into one-dimensional array of pixels. means for decoding the decoded pixel data into data, means for temporarily storing the decoded pixel data, means for obtaining information on the position of non-zero pixel data within the block at the time of storage, and means for temporarily storing the information on the position of the non-zero data. The pixel data is read out from the pixel data temporary storage means and the read pixel data is output-controlled; A device that obtains information on the position of pixel data, temporarily stores it, and reads out the temporarily stored pixel data.
In the above method, if the block to be encoded or decoded has consecutive zero coefficient data at the end of zigzag scanning, the encoding is performed without scanning the last consecutive zero coefficient data of the block. It is possible to end the processing, and even in decoding, it is possible to end the processing without reproducing the last continuous zero coefficient data of the block. If the block has continuous zero coefficient data at the end of the zigzag scan, the processing in the encoding and decoding processes can be reduced to improve the efficiency of the overall processing. An encoding device according to an embodiment of the invention is shown, including pixel data processing means for processing input pixel data, means for temporarily storing pixel data to be encoded, and pixel data to be encoded with zero power Q. a means for counting at least one or more non-zero number of pixel data to be encoded; a means for temporarily storing the counted value; and a means for temporarily storing the pixel data to be encoded. It consists of a means for reading into an array, a means for comparing count values, and a means for encoding pixel data to be encoded. A block diagram of one embodiment is shown. In the basic block of the encoding device shown in FIG. 6, FIG. 1 corresponds to the encoding process of the encoder/decoder 66. FIG. 1 (1) shows the scanning of coefficient data. Figure 1 (2) is a block diagram for reading data from the buffer and encoding it. 1 is the coefficient data or quantum data after DCT processing. The coefficient data after conversion 2 is the coefficient data transfer control signal Jii! P, 3 is an address generation a to the buffer memory; 4 is an input data system that controls input coefficient data; 5 is a call register; 6 is a count register that temporarily stores the counter value; 7 is a buffer that temporarily stores the coefficient data; 8 is a non-zero data detection signal 9 is the counter from the register & 10 is the address to the buffer memory 11 is the block end signal 12 is the coefficient data to be encoded 13
The count comparator 14 is the counter, the data comparator 16 is the count value 17 from the count register 6, and the count match signal 18 is the count from the counter &19
is a non-zero data detection signal! Coefficient data from buffer 20 Address 22 to buffer 21 indicates code image 23 is code data, coefficient data 1 is input to input data control 4, and the coefficient data is data after DCT processing (dai quantization processing) If the data is the result of quantization processing, then
The data as it is is input to the buffer 7, and then it is determined whether or not the data is "zero" except for the DC component coefficient (coefficient data at the top left in the block), and as a result, it is determined that it is "not zero". Then, if the non-zero data detection signal 8 is used, the number of non-zeros in the block is counted by the counter 5.The input timing of the input coefficient data is obtained by the transfer control signal 2, and the buffer is In many cases, the address given to 7 is an address that scans in the horizontal or vertical direction as shown in Figure 3 (2>, (3)).
When one block of coefficient data 1 is transferred, the count value of the counter 5 at that time is stored in the count register 6.
-1 block of buffer input is completed Buffer 7
The input data is converted into a one-dimensional array and encoded by the address corresponding to the zigzag scan as shown in Fig. 3 (1) output from the encoder 22. The output data excluding the coefficient of the DC component is judged by a data comparison 15 to determine whether it is 'zero' or not. If the result is determined to be 'not zero', a non-zero data detection signal 18 is output and the counter 14 counts the number of times. The encoder 22 combines the data from the buffer 7 with the run length value of zero data and the following non-zero data value to generate Huffman encoded data.
Even if it is output as code data 23, the count comparison 13 compares the count value 18 of the counter 14 and the count value 16 of the count register 6, and if they match, it notifies the encoder 22 of the count match signal 17. 2
2 stops the code processing based on that information.If all the processing in the block has not been completed at that point, add the UEOB code.If all the processing in the block has been completed, it will end as is. If we use the encoding process,
In the encoding process after one-dimensional array conversion, if the last data in the block is "zero", the encoding process can be performed without scanning that data.
To explain this in the figure, non-zero coefficient data excluding the DC component is represented by (X, Y) coordinates. When shown in the horizontal scanning order of Figure 3), (1, 0), (5, 0) , (0,
1), (5, 2), and (0, 3). If only the coefficient data of this AC component is encoded by zigzag scanning, the encoded data up to the coefficient of (1, 0) C (0,
1), code data up to (0,1) C(0,2), (0
, 1), code data C (5, 1) up to (5, 0), code data C (14, 1) up to (5, 2), and this encoding At the time of processing, five pieces of non-zero coefficient data were detected, so we connected the EOB code and finished the encoding process, but when we actually looked at the coefficient data that was not scanned after that process, all of them had a value of zero. However, it is not clear that the obtained encoded data is correct.In this way, compared to the conventional example, it can be seen that the processing of consecutive zero data at the end of zigzag scanning within a block is more efficient.
In addition, in the present invention, by using a plurality of count registers 6 and buffers 7 in Figs. It can be done and the efficiency can be further improved.

Next, a decoding device according to an embodiment of the present invention will be described.

Means for decoding encoded data into pixel data in a one-dimensional array, means for temporarily storing the decoded pixel data, means for obtaining information on the position of non-zero pixel data within a block at the time of storage, and the position of the non-zero data. The system is comprised of means for temporarily storing information, means for reading out pixel data from the temporary storage means for pixel data, and means for performing output control on the read out pixel data. Figure 2 shows a block diagram of an embodiment of the present invention. Figure 2 (1) shows the scanning of decoded data until it is stored in a buffer for conversion. Figure 2 (2) shows the buffer. 31 is a code data register 32 is a decoder 33 that compares decoded coefficient data; 34 is a reproduction position register 35 that stores the reproduction position of a coefficient within a block; is a coefficient position register for temporarily storing information in a playback position register; 36 is a buffer for temporarily storing coefficient data;
37 is the decoded coefficient data 38 is the address given to the buffer 39 is the data processing end signal 40 in the block
A non-zero data/event detection signal 41 is a non-zero data playback position data of the block. 42 is a control for outputting coefficient data. An output data control 43 is an address generator for giving an address to the buffer. An address generator 44 is a data generator for each coefficient data. 45 is the coefficient data from the buffer. 46 is the address from the address generator. 47 is the transfer control signal that controls the output timing of the output data. The transfer control signal 48 is the coefficient data to be output. , 31 is decoded into one-dimensional array data by the decoder 32 and transferred to the buffer 36. At that time, the address 38 corresponding to the zigzag scan as shown in Fig. 3 (1) is assigned to the buffer. Even if the data is given in synchronization with the decoded data, the decoded coefficient data excluding the DC power coefficient is simultaneously input to the comparator 33, and it is determined whether it is zero or not (L non-zero data). Output the detection signal 40, input the coefficient data 37 and the address 38 synchronized with the non-zero detection signal 40 to the playback position register 34, and store the position of the non-zero data. is always a non-zero data position. When the decoder 32 detects the end of a block, it outputs a data processing end signal 39 in the block, and the reproduction position register 34 outputs non-zero data reproduction position data 41 of the block. The data is temporarily stored in the coefficient position register 35, and at the same time, the data in which at least the AC component is located in the reproduction position register 34 is cleared. (2) or (3), the address generator 43 obtains the output timing using the transfer control signal 47 and controls the address 46 to the buffer. register 35
These data (
Coefficient data 45. Non-zero detection data 44) is input to the output data control 42, and if each coefficient is non-zero data, if the subsequent processing is inverse quantization processing, the coefficient data 45 is input as is.
If it is inverse DCT processing, it will output data that has undergone inverse quantization processing, and if it is not non-zero data, it will mask the numbered coefficient data and output it as zero. ○If the block ends with a B code, the decoding process for that block ends when the EOB code is detected, and by controlling the output of coefficient data, the last continuous zero data of the block is However, the processing of one embodiment of the present invention will be explained using the example of actual encoding processing shown in FIG. 4. However, when the encoded data shown in FIG. Only the part scanned by the solid line in the block shown in (1) is written to the buffer, and the coefficient data part scanned by the broken line is not written to the buffer.1+~ Therefore, the previous block does not write to that part. If you are using ICE, the data in the buffer that has not been scanned is not necessarily zero.
As shown in Figure 4 (3), the contents of the coefficient position register indicate the coefficient position of the actually scanned coefficients and non-zero data.
Since the coefficient position register in Figure (3) is only the coefficient data with data of 1'', as a result, the coefficient data of the scan in the broken line area in Figure 4 (1) is output as zero. In this example, if the code data ends with an EOB code, correct decoded data can be obtained without scanning the last continuous zero coefficient data part. , G!
Figure 2 (IL (2)) Situation where multiple coefficient position registers 35 and buffers 36 are used alternately Is it possible to perform the processing shown in Figure 2 (1) and (2> in a pipeline manner? The efficiency can be further improved with the encoding device of the present invention: When the coefficient data of at least one or more consecutive zeros at the end of a block is encoded, the consecutive zeros are processed. The decoding process of the present invention also makes it possible to finish the encoding process without scanning the coefficient data of ↓ EOB.
When decoding code data connected by codes, it is possible to perform the decoding process without scanning the last continuous zero coefficient data in the block.

[Brief explanation of the drawing]

Fig. 1 shows a block of an encoding device in an embodiment of the present invention; Fig. 2 shows a block of a decoding device in an embodiment of the invention; Fig. 3 shows an example of scanning data within a block; Figure 5 shows an example of actual encoding processing. Figure 6 is a block diagram showing a conventional encoding device and decoding device. 1... Input coefficient data 2... Transfer control double image 3 ...Address generator 4...Manual data control 5...Karan Hisashi 6...Count register 7
...Buffer, 8...Non-zero data detection magnification 9...
・Counter t 10...Address to buffer 1
1...Block end signal q 12...Coefficient data to be encoded 13...Count comparison 14...
Karan Hisashi 15... Data comparison 1 & 16... Count summary from the count register 17... Count - Shinshin Jfi! P, 18... Count 19 from the counter
. . . Non-zero data detection signal 20 . . . Coefficient data from the buffer 21 . . Address input to the buffer 22 . .
・Encoder 23... Code data 3 engineering... Code data 32... Decoding method 33... Coefficient data comparison kernel 34.
...Playback position Regis ku 35...Coefficient position Regis ku 3
6...Buffer, 37...Decoded coefficient data
38...Atrear given to buffer, 39°...Data processing end signal in block 40...Non-zero data detection signal 41...Non-zero data reproduction position data of block 42...Output data control 43 ...Address generator 44...Non-zero detection data 45...Coefficient data from the buffer 46...Atrea S47 from address generator 47...Transfer control signal 48...Coefficients to be output data.

Claims (6)

[Claims] (1) Divide the image data into N×M blocks (N and M are natural numbers), perform orthogonal transformation and quantization processing on the blocks, and convert the data into one-dimensional array data,
In the encoding processing device, pixel data processing means for processing input pixel data, means for temporarily storing pixel data to be encoded, means for determining whether or not the pixel data to be encoded is zero; means for counting the number of at least one or more non-zero pixel data to be encoded, means for temporarily storing the counted value, and means for reading out the pixel data to be encoded from the temporary storage means into a one-dimensional array; It has a means for comparing count values and a means for encoding pixel data to be encoded, and when the pixel data to be encoded is stored in the temporary storage means, the input pixel data is encoded by the pixel data processing means. Obtain the pixel data to be processed and the information that the pixel data is non-zero,
The counted value is temporarily stored, and when the temporarily stored pixel data to be encoded is converted into a one-dimensional array and read out, the number of non-zeros in that data is counted again, and the temporarily stored non-zero number is An encoding device characterized by performing encoding termination processing when a count value is reached. (2) The encoding device according to claim 1, wherein the pixel data to be encoded in the pixel data processing means is input pixel data. (3) The encoding according to claim 1, wherein the pixel data processing means performs a quantization operation on the input pixel data, and uses the operation result as the pixel data to be encoded. Device. (4) Divide the image data into N×M blocks (N and M are natural numbers), perform orthogonal transformation and quantization processing on the blocks, and convert the data into one-dimensional array data,
In an encoding processing device, means for decoding encoded data into pixel data in a one-dimensional array, means for temporarily storing the decoded pixel data, and means for obtaining information on the position of non-zero pixel data within a block at the time of storing the decoded pixel data. and means for temporarily storing information on the position of the non-zero data, means for reading pixel data from the temporary storage means for pixel data, and means for applying output control to the read pixel data, and the pixel data is decoded. When the stored pixel data is stored in the temporary storage means, information on the position of the non-zero pixel data is obtained and temporarily stored, and when the temporarily stored pixel data is read out, the position of the stored non-zero pixel data is obtained. A decoding device characterized by controlling output pixel data based on information. (5) The means for controlling the output of pixel data uses the information of the position of non-zero pixel data to output the pixel data as is if it is non-zero, and to output the pixel data as zero otherwise. The decoding device according to claim 4, characterized in that: (6) The means for applying output control to pixel data uses information on the position of non-zero pixel data, and if it is non-zero, performs inverse quantization processing, otherwise outputs pixel data as zero. 5. The decoding device according to claim 4, characterized in that: