JP2000165873A - Moving image information compression method and system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance image quality and sound quality by eliminating the need for block conversion so as to apply compression processing to data possible for prediction coding of image information at high compression rate and at high speed. SOLUTION: The system is provided with an entropy coding means that reduce inter-frame redundant information by comparing pixels adjacent spatially and temporally to each other so as to output difference information, with a bit map information recording means 4 that sequentially compares inter-frame pixel t with pixel t-1 and stores information where an absolute value of the difference is larger than a given parameter P and other information to a 1-bit bit map and with an information compression means 5 that applies processing to the information where an absolute value of a difference between both the pixels t, t-1 stored in the bit map information recording means 4 is not larger than the parameter P as inter-frame pixels having no change and applies compression processing to the information where the absolute value of the difference between both the pixels is larger than the parameter P.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and a system for compressing moving image information for performing high-speed compression processing on a moving image information amount at a high compression ratio.

[0002]

2. Description of the Related Art Conventionally, it has been customary to convert an image signal into another signal once, and then perform an coded transmission by assigning an appropriate code by utilizing the statistical properties of the converted signal. It is. In this case, in the redundancy within one frame, for example, in an image of a regular pattern or a flat image, since the correlation between adjacent pixels is strong, the value of the pixel to be encoded next is reduced to some extent from the value of the already encoded pixel. The so-called predictive encoding is performed, in which only information that can be predicted and that cannot be predicted is extracted and coded to perform significant information compression.

[0003] For example, in a moving image such as a videophone,
Images of successive frames are very similar and temporal changes are often limited, and such temporal redundancy can be removed by inter-frame prediction coding using prediction over frames. Have been. At this time, in general, a block code in which one code word is assigned to one symbol is adopted, one frame is divided into smaller pixel blocks, and the property that the difference in luminance is reduced in each block is reduced. A so-called block encoding process, which is utilized for information compression, is employed.

Further, a so-called entropy coding for realizing data compression by allocating an efficient code to a converted signal and a Huffman coding method as an efficient code generation method are known. A typical example is an arithmetic code, which divides a probability number straight line according to the appearance probability of a symbol sequence and uses a binary decimal value indicating a position in the divided section as a code for the sequence. That is, code words are sequentially formed by arithmetic operations.

In the prior art, a three-step block coding system for efficiently coding an image signal consists of sampling, conversion and quantification. In order to maintain the planar resolution and high-frequency components of the image signal at this time, it is usually necessary to perform sampling at twice the speed of the highest frequency component at that frequency.

[0006]

However, in the conventional image signal information compression method, since a complicated block coding method is used, data capable of predictive coding of image information is compressed at a high compression rate and at a high speed. It was difficult to perform compression processing.

In the moving image compression processing, when normal difference information is compressed, that is, when A1 and A2 are expected to be close to each other and the value of A1 can be known before the information of A2, A2-A1 is calculated. Since a conventional compression method using a Huffman code, an arithmetic code, or the like is adopted on the assumption that the occurrence probability near 0 is high, the possible values of A1 and A2 are 0.
Ｎn, the possible range of the difference A2-A1 is 2n +
One way, 2n + 1 kinds of Huffman code words must be prepared. However, there are actually n possible values of A2, and the n codes are not used when viewed locally, so that there is a problem that the code becomes redundant accordingly.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described conventional problems, and has been developed in view of a moving image which can perform compression processing at high compression rate and high speed on data capable of predictive coding of image information. It is an object of the present invention to provide a method and a system for compressing information.

[0009]

For this reason, in the moving picture information compression method according to the present invention, spatially and temporally adjacent pixels are compared with each other, and the difference information is output so that the inter-frame information is output. Is a method of compressing moving image information to reduce redundant information, sequentially comparing each pixel between the frames, the part whose absolute value of the difference is greater than a given parameter and the other part Is stored in a 1-bit bitmap, the portion where the absolute value of the difference is not larger than the parameter is processed as a pixel that does not change between frames, and the information of the other portions larger than the parameter is subjected to compression processing. Has solved the above-mentioned problem.

In the moving picture information compression system according to the present invention, spatially and temporally adjacent pixels are compared with each other, and difference information is output to reduce redundant information between frames. What is claimed is: 1. A video information compression system comprising entropy coding means, wherein a pixel t and a pixel t-1 between said frames are sequentially compared, and a portion where the absolute value of the difference is larger than a given parameter P Bitmap information recording means for storing the data and other parts in a 1-bit bitmap, and both pixels t and t-1 stored by the bitmap information recording means.
The portion where the absolute value of the difference is not larger than the parameter P is processed as a pixel that does not change between frames, and the information of the other portions that are larger than the parameter P is compressed by having an information compression unit that performs a compression process. The above-mentioned problem has been solved.

Further, the 1-bit bitmap information stored by the bitmap information recording means is run-length, modified READ (MR,
MMR), modified Huffman (MH), JBIG, and the like, the information is compressed to solve the above-described problem.

The information compression means for compressing information of a portion larger than the parameter P employs an adaptive Huffman coding process having a plurality of Huffman tables for prediction information, thereby solving the above-mentioned problem.

[0013] In addition, the entropy encoding means has an arithmetic table for predictive information, and adaptive arithmetic encoding processing for selecting and encoding a plurality of tables based on the predictive information is added. Also, the above-mentioned problem has been solved.

The basic structure of the method and system for compressing moving picture information according to the present invention is to compare spatially and temporally adjacent pixels (pixels) and output difference information, thereby providing redundant information between frames. Is to reduce. That is, the pixel t and the pixel t-1 between the frames are sequentially compared, and the portion where the absolute value of the difference is larger than the given parameter P and the other portion are recorded as 1-bit bitmap information. It is.

In the entropy encoding process, redundant information is reduced by predicting a code appearing between frames and outputting a slight deviation from the prediction. Other information is compressed by an adaptive Huffman coding method or an adaptive arithmetic coding method in which a plurality of tables are selected and coded based on prediction information.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of the encoder compressor. For example, NTSs such as video cameras, disc players or video cassette players
A composite analog signal output from the C-type apparatus is converted into a digital signal by an analog-digital converter 1, digitally output as representing one line of a video frame, and stored in a buffer 2.

As shown in FIGS. 1 and 2, the encoder compressor sequentially compares a pixel t and a pixel t-1 between frames, and obtains the absolute value of the difference from a parameter P given. A bitmap information recording means 4 for storing a portion having a larger size and other portions in a 1-bit bitmap is provided. Moreover, both pixels t, t stored by the bitmap information recording means 4 are stored.
For example, a part having an absolute value of the difference of −1 not larger than the parameter P is processed as a pixel having no change between frames, and information of a part larger than the parameter P is subjected to compression processing. An information compression means 5 is provided by adaptive Huffman coding processing having a table for prediction information. Then, the encoder compressor compares, for example, an arithmetic table for reducing redundant information between frames by comparing spatially and temporally adjacent pixels (pixels) and outputting difference information. And entropy coding means 6 for adding an adaptive arithmetic coding process for selecting and coding a plurality of tables based on the prediction information.

After the encoding, the pixel data for each frame is stored in the memory 1 of the bitmap information recording means 4.
0, where it is delayed by one frame time and exists as the immediately preceding frame, and then filtered by the temporal filter 11, which is a time variable impulse response filter. After filtering, the current frame memory 13 and the previous frame memory 14 are checked for redundancy between frames by a compressor and the difference is calculated. That is, each pixel coded by the comparing means 12 is compared with the corresponding pixel of the immediately preceding frame. Each pixel is marked with a single bit that defines whether it is a new pixel or unchanged from the previous pixel. This process creates a frame bitmap with one bit per pixel. At this time, the bitmap for each frame is distinguished by comparison between the frames.

The 1-bit bitmap information stored by the bitmap information recording means 4 is run-length, modified READ (MR,
MMR), modified Huffman (MH), JBIG and the like. Specifically, in the case of run-length encoding, in general, a binary document image handled by facsimile or the like often has a certain number of white pixels or black pixels that appear to some extent, and white or black is continuous in one-dimensional direction. A so-called run, which is a lump of pixels, is used as an encoding unit, and the length of the number of consecutive pixels is encoded as a run length. For example, in a digital facsimile using a public telephone network, it is customary to use a modified Huffman code configured separately for black and white for a run length model.

In the case of modified Huffman coding (MH), it is read at an image density of 8 pixels / mm, for example.
It is adopted as a one-dimensional encoding method for facsimile transmission for obtaining black and white pixel information of 1728 pixels per scanning line, and the MH code is a continuous mass of white pixels (white run),
Alternatively, it expresses the run length, which is the length of a block of black pixels (black run), and uses the fact that the probability of occurrence of a white run or black run of a certain length is statistically biased to use a variable-length code. Assigning is based on the principle of data amount compression.

In the case of the modified READ (MR, MMR), this is adopted as a standard of the two-dimensional encoding system in addition to the one-dimensional encoding system. After encoding, up to one at standard resolution,
This is for two-dimensional encoding of up to three continuous scanning lines at high resolution. In the case of MMR, the MR encoding method is set to infinity for both the standard resolution and the high resolution.

The basic structure of the method and system for compressing moving picture information according to the present invention is to compare pixels spatially and temporally adjacent to each other and to output difference information to obtain redundant information between frames. Is to reduce. That is, as shown in FIG. 2, the pixel t between pixels and the pixel t-1 are sequentially compared, and a portion where the absolute value of the difference is larger than a given parameter P and a portion other than the parameter P are 1 bit. Is recorded as bitmap information. Then, the entropy coding unit 6 predicts a code appearing in each of the frames and between the frames, and outputs a slight deviation from the prediction to reduce redundant information. At this time, it is well known that, when coding transmission is performed after coding allocation, the average code length per pixel does not become less than the average information amount (entropy).

The algorithm of the adaptive Huffman coding method will be described below. This adaptive Huffman code is intended to collectively perform a series of predictive encoding processes of difference information generation and Huffman encoding to improve codeword efficiency. In the conventional Huffman coding process, a codeword is generally generated using one Huffman table, and 1 code is generated in a dynamic Huffman code.
The Huffman table is updated every time a word is encoded, but the Huffman table (code table) is used in adaptive Huffman coding.
And a plurality of tables are selected and encoded by a table selector based on the prediction information. As a result, data that can be predicted and encoded, such as image information and audio information, can be efficiently compressed.

The algorithm of the adaptive arithmetic coding method will be described below. This adaptive arithmetic code processes a series of predictive encoding processes of difference information generation and arithmetic encoding collectively,
This is for improving the efficiency of the code word. In the conventional arithmetic coding process, a code word is usually generated using one occurrence probability table, and the occurrence probability table is updated each time one word is coded in the dynamic arithmetic code. An arithmetic table (decoding table) is provided for prediction information, and a plurality of tables are selected and encoded by a table selector based on the prediction information. As a result, data for which predictive coding of image information is possible can be efficiently compressed.

As the adaptive transform coding for the moving image, a method is employed in which the transform coefficients are scaled by feedback control using the sufficiency of the transmission path buffer memory and then coded. In this case, a threshold for determining an insignificant coefficient not to be encoded based on a histogram of transform coefficients for a representative image, a run-length code for encoding the continuity of the insignificant coefficient, and an adaptive Huffman code table for encoding a significant coefficient are obtained. , An encoding method based on this is adopted. As shown in FIG. 3, a specific configuration of the predictive encoding circuit is such that analog-to-digital converted image input data is delayed halfway and the value of the previous data (the data with the optimal delay) is stored in a table. At the same time as being sent to the selector and encoded, the image input data is compared with the data directly transmitted to the encoder, and the difference is obtained. In the table selector, a code word is selected from the code table based on the prediction information according to the image input data and sent to the encoder, where the image input data is converted into a code word adjusted by information compression. As shown in FIG. 4, the specific structure of the predictive decoding circuit is such that the codeword is transmitted to the decoder, and at the same time, the directly transmitted codeword is once sent to the table selector, where it is predicted from the decoding table. The image input data is selected based on the information and sent to the decoder to obtain a difference from the already decoded pixel value to obtain image input data adjusted.

[0026]

The present invention is configured as described above.
In particular, by eliminating the block conversion, it is possible to perform high-speed compression processing on data capable of predictive coding of image information at a high compression rate and improve image quality and sound quality. In addition, adaptive Huffman compression processing and adaptive arithmetic compression processing collectively process conventional differential information generation and Huffman coding, or a series of predictive encoding processing of differential information generation and arithmetic coding, thereby improving codeword efficiency. It is possible,
Data capable of predictive coding of image information can be efficiently compressed.

The one-bit bitmap information stored by the bitmap information recording means is run-length, modified READ (MR, M
MR), modified Huffman (MH), JBIG, etc., the data is compressed, so that data capable of predictive coding of image information can be efficiently compressed. For example, image quality can be reduced in a digital facsimile using a public telephone network. Can be improved.

Further, the information compressing means for compressing the information of the portion larger than the parameter P is based on the adaptive Huffman encoding process having a plurality of Huffman tables for the prediction information. Can be selected and encoded by a table selector on the basis of the data, thereby enabling efficient compression of data capable of predictive coding of image information.

The entropy encoding means includes:
Arithmetic tables are divided into prediction information, and adaptive arithmetic coding processing to select and encode multiple tables based on prediction information is added. Thus, data that can be predicted and encoded for image information can be efficiently compressed.

[Brief description of the drawings]

FIG. 1 is a block diagram of an encoder compressor according to an embodiment of the present invention.

FIG. 2 is a block diagram similarly.

FIG. 3 is a circuit diagram illustrating an encoding circuit.

FIG. 4 is a circuit diagram illustrating a decoding circuit.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Analog-digital converter 2 ... Buffer 3 ... Encoder 4 ... Bitmap information recording means 5 ... Information compression means 6 ... Entropy coding means 10 ... Memory 11 ... Temporal filter 12 ... Comparison means 13 ... Current frame memory 14 ... Previous frame memory

Claims (5)

[Claims] 1. A moving image information compression method for comparing spatially and temporally adjacent pixels and outputting difference information to reduce redundant information between frames.
The respective pixels in the frames are sequentially compared with each other, and a portion where the absolute value of the difference is larger than a given parameter and a portion other than the given parameter are stored in a 1-bit bitmap. A method for compressing moving image information, characterized in that a portion that is not large is processed as a pixel that does not change between frames, and information of other portions that are larger than a parameter is subjected to compression processing. 2. A moving image information compression system comprising an entropy coding unit for comparing spatially and temporally adjacent pixels and outputting difference information to reduce redundant information between frames. Then, the pixel t and the pixel t−1 between the frames are sequentially compared,
Bitmap information recording means for storing a portion having an absolute value of the difference larger than the given parameter P and the other portion in a 1-bit bitmap; and both pixels t stored by the bitmap information recording means. , T-1 are processed as pixels that do not change between frames, and the other parts of the information larger than the parameter P are compressed. A compression system for moving image information, characterized in that: 3. The 1-bit bitmap information stored by the bitmap information recording means is run-length, modified READ (MR, MM
3. The moving image information compression system according to claim 2, wherein the information is compressed by a method such as R), modified Huffman (MH), JBIG, or the like. 4. The moving picture according to claim 2, wherein the information compression means for compressing information of a portion larger than the parameter P is based on an adaptive Huffman coding process having a plurality of Huffman tables for prediction information. Image information compression system. 5. The method according to claim 1, wherein said entropy coding means has an arithmetic table for predictive information, and adaptive arithmetic coding processing for selecting and coding a plurality of tables based on the predictive information. 3. A moving picture information compression system according to claim 2, wherein: