KR0134324B1 - Variable length method of data compression - Google Patents

Abstract

The present invention relates to a method of encoding digital image data, wherein the image data of N-bit units quantized by coefficients in a frequency domain is classified for each bit plane unit from MSB to LSB and is variable in consideration of characteristics of each bit plane. The present invention relates to a variable length coding method capable of performing optimal image data compression by applying a long coding table for each bit plane.

Description

Bitwise Variable Length Encoding Method for Data Compression

1 is a block diagram showing an example of a conventional video data encoding apparatus.

2 is an explanatory diagram for explaining a conventional variable length encoding method.

3 is a conceptual diagram for explaining an encoding method of the present invention.

4 is a flowchart of a variable length coding method according to the present invention.

5 is an exemplary diagram of the i th bitplane of FIG.

6 is a schematic diagram each showing examples of the scanning method of the present invention.

The present invention relates to a method of encoding digital image data, and more particularly, to a method of variable length encoding in consideration of characteristics of each bit plan in a method of receiving variable length encoding for lossless data compression by receiving data in units of N bits. will be.

In general, in a system for transmitting and receiving video and sound, a video signal and an audio signal are encoded and transmitted as digital signals or stored in a storage unit, and then decoded and reproduced. Methods used for encoding such video signals include a transform encoding method, a differential pulse code modulation method, a vector quantization method, and a variable length encoding method. These coding methods are used to compress the total amount of data for removing redundancy data included in the digital video signal.

FIG. 1 is a block diagram showing a general digital signal encoding apparatus. The method of transforming a KxK block by DCT (Discrete Cosine Transform) is used to quantize transform coefficients and to perform variable length encoding on quantized data. Means for further compressing the data amount, and means for performing inverse compensation by inverse quantization and inverse transformation of the quantized data.

In FIG. 1, an image signal input through the input terminal 10 is converted into a signal in the frequency domain in units of K × K blocks by the K × K converter 11, and the energy of the converted conversion coefficient is mainly directed toward the low frequency side. Are gathered. For each block, data conversion may use methods such as a Walsh-Hadamard Transform (WHT), a Discrete Fouier Transform (DFT), and a Discrete Sine Transform (DST). The quantization unit 12 converts the transform coefficients into representative values of a predetermined level. The variable length coding unit 13 further compresses the data by variable length coding using the statistical characteristics of the representative values.

On the other hand, the quantization step size (QSS), which is converted according to the state of the buffer 14 in which the variable length coded data is stored, controls the quantization unit to adjust the transmission bit, and is also transmitted to the receiving side for use in the encoding apparatus.

In addition, since there are many similar parts between screens in general, when the screen is in motion, the motion vector (MV) is calculated by estimating the motion, and when the data is compensated using the motion vector, the difference signal between adjacent screens. Is very small, which can further compress the transmission data. In order to perform such dynamic compensation, the inverse quantization unit 15 and the K × K inverse transform unit 16 inversely quantize the quantization data output from the quantization unit 12 and inversely transform the image signal in the spatial domain. The video signal output from the inverse transform unit 16 is stored in frame units in the frame memory 17, and the synchronization unit 18 stores the K × K block data of the input terminal 10 in the frame data stored in the frame memory 17. The block having the most similar pattern is found to calculate a motion vector (MV) representing the motion between the two blocks. This motion vector is transmitted to the receiving side, used in the decoding apparatus, and transmitted to the upper eastern part 19. The dynamic compensator 19 receives a dynamic vector from the dynamic weighting unit 18, reads a K × K block corresponding to the dynamic vector MV from previous frame data output from the frame memory 17, and then inputs the dynamic vector. Supply to the adder (A1) connected to. Then, the adder A1 calculates the difference between the K × K block supplied to the input terminal and the K × K block of the similar pattern supplied from the dynamic compensator 19, and the output data of the adder A1 is as described above. It is encoded by the process and transmitted to the receiving side. That is, the entire video signal is first transmitted and only the difference signal due to the movement is transmitted afterwards.

On the other hand, the data whose motion is compensated for in the compensator 19 is added to the video signal output from the KxK inverse transform unit 16 in the adder A2 and then stored in the frame memory 17. The refresh switch SW is turned off from time to time by other control means so that an input video signal is encoded and transmitted in a PCM mode, thereby accumulating encoding errors due to encoding and transmitting only difference signals. Refresh at time intervals and allow the receiving side to deviate from the transmission error on the channel within a certain time.

The encoded video data is transmitted to the receiver and output as a video signal through a decoding apparatus.

In the conventional video data encoding system as described above, the data converted by DCT, WHT, DFT, DST, etc. is variably encoded in block units of a predetermined size. An example is described with reference to FIG.

2 illustrates an example of transformed image data of a block having a predetermined size. As shown in the drawing, the component values of the image data may be the same or may be different from each other. When the component values in this block are arranged in a one-dimensional column by zigzag scanning in the order indicated by the arrows, 150, 30, 0, 0, 20, 0, 10, 0, 0, 25,... … Becomes Here, the run-length of 0 can be calculated and expressed as a pair of run and amplitude (0,150), (0,30), (2,20), (1,10), (2,25),... … Becomes The sign corresponding to this (run, mpretude) pair can be obtained, and variable length coding can be performed for the entire block. Of course, the Huffman table should be prepared in consideration of the characteristics of the block.In general, a short length encoding is assigned to a pair with a high probability of occurrence and a long length code is assigned to a pair with a low probability of occurrence. It is to be reduced. Using the above method, the converted image data can be extruded to some extent. However, in the variable length coding method described above, when each component of the converted image data as shown in FIG. Up to, each of the bit planes of the same level of bits (Bit Plane) is different from the appearance statistical properties appearing 0, 1, there is a drawback that can not fully exploit the difference.

Accordingly, an object of the present invention is to efficiently utilize the appearance statistical properties of 0 and 1 that appear differently on each bitplane by variably length coding each bitplane unit from MSB to LSB of data to be compressed. The present invention provides a variable length coding method for enabling variable length coding.

The object of the present invention is to convert the quantized image data having a predetermined block size into a transform coefficient of the frequency domain, and then scan in a predetermined direction to variably code the length of each bit plane. Preparing a variable length encoding table in consideration of statistical properties, dividing the quantized data into binary bits, dividing the data into N bitplanes, and scanning the bitplanes to perform a pair of zero runs and one ampli- tude. And a variable length encoding method using the variable length encoding table.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

3 shows binary values of component values of image data in three-dimensional spatial yarns. Each data column in the vertical direction having K component coefficients is a vertically scanned image data of a block having a certain size by a zigzag scan method, and each component is displayed in binary bits. That is, binary data is obtained by scanning image data of M blocks by a predetermined scan method. Each data component consists of N bits, and each bit from the MSB to the LSB gathers to form N bit planes.

The size of the bit N is determined by the number of bits of the largest component value, and the remaining bits of the component value composed of bits smaller than N bits are composed of zero bits. For example, if the largest component value of the image data is 150 and one component value of the image data is 30, N is 8, '150 is represented by 10010110, and 30 is represented by 11110.

Each bitplane is represented by 0 and 1 only, and the appearance statistics of 0 and 1 are different. In general, a bitplane on the MSB side has a higher probability of appearing than '1' than a bitplane on the LSB side, and there may be a bitplane having similar appearance statistics.

Therefore, if variable length coding is prepared independently for each of the N bit planes composed of the N bits from the MSB to the LSB of the image data, and the variable length coding is performed for each bit plane, more data compression is performed than in the conventional method. It can be seen that. A variable length encoding table for each bitplane is created by considering the appearance statistics of 0 and 1 of each bitplane. That is, when scanning by one scanning method selected for each bitplane, since each bitplane is often different in appearance statistics of 0 and 1, the statistical properties of the run length of 0 'will be different. Therefore, a variable length encoding table is created by considering the statistical properties of zero runlengths for each bitplane. Of course, in the processing order of each bit plane, processing may be performed sequentially from the MSB side to the LSB, sequentially from the LSB side to the MSB, or in another arbitrary order.

The flow of the variable length coding method for each bitplane is shown in FIG. The order in which the bitplanes are scanned is determined for the N bitplanes (step 40) of the quantized data (step 41), and the bitplanes to be first variable-length coded are designated (step 42). Accordingly, the variable length encoding table 46 corresponding to each bitplane is used to sequentially specify the bitplanes to be variable-coded after the variable-length encoding for the first bitplane (step 47), and then start from the first bitplane. Variable-length encoding is executed in sequence up to the last bitplane (step 43). In this case, after variable length encoding on the corresponding bit plane, it is determined whether the bit plane is the last bit plane to be variable length coded (step 44), and if the bit plane is not the last bit plane to be variable length coded, If the bitplane is repeated and the bitplane is the last bitplane to be variable-coded, the variable-length encoding process for all bitplanes is terminated (step 45). Variable length encoding of any i th bitplane in the above flow chart will be described in more detail.

5 shows the i-th bitplane. In FIG. 5, the i-th bitplane has a size of K × M having K blocks vertically and M blocks horizontally. The scanning method selected for this bit plane is arranged in one-dimensional bit strings. Examples of how to scan a bitplane are shown in FIG. FIG. 6A illustrates a method of scanning bit by row of a bitplane, and FIG. 6B illustrates a method of scanning bit by column of a bitplane. The scanning method of the bitplane is not limited to the method shown in FIG. 6 (a) and 6 (b) but other methods may be used. If the sequence of bits is rearranged into a pair of zero run and one mpretude of the bit string, and the code corresponding to the pair is found on the variable-length encoding table prepared by considering the appearance statistical characteristics of the 0 and 1 of the i-th bitplane, i Variable length coding for the first bitplane can be performed.

Of course, when there are several bit planes with similar appearance statistics of 0 and 1, a plurality of bit planes may be processed in small units using a common encoding table.

As described above, the variable length coding of each bitplane from the MSB to the LSB is separated and encoded in each bitplane unit so that the optimal variable length coding can be performed because the statistical appearance characteristics of 0 and 1 for each bitplane can be sufficiently used. . In addition, since the variable length encoder needs to consider only 0s and 1s on the bitplane, there is an effect that it can be simplified.

Claims (4)

A method of converting image data having a predetermined block size into a transform coefficient of a frequency domain, quantizing, scanning in a predetermined direction, and variable length encoding, wherein the variable length is considered in consideration of the appearance statistics of 0 and '1 on each bitplane. Preparing an encoding table, displaying the amplified data as binary bits, dividing the data into N bitplanes, scanning the bitplanes, and displaying them as a pair of zero runs and an '1' ampli- tude; and And variable length encoding the bit plane using the variable length encoding table. The variable length coding method of claim 1, wherein the variable length coding is performed for each bit plane. The variable length coding method of claim 1, wherein a common variable length coding table is prepared for the bit planes having similar appearance statistics of zeros and ones. 4. The variable length encoding method according to claim 3, wherein common bit planes are grouped and variable length encoded using the common encoding table.