CN116527904B - Entropy coding method, entropy decoding method and related devices - Google Patents

Abstract

This application provides an entropy coding method, an entropy decoding method and related devices, which are applied in the field of artificial intelligence, including: performing unsigned fixed-length coding on quantized data to obtain an unsigned fixed-length code; determining based on prediction residual data bit indication information and asymmetric value range flag bits to obtain the transmission value range to determine the encoding method; encode the prediction residual data according to the encoding method to obtain a signed fixed-length code; encode only the quantized data to obtain an unsigned variable Long code; encode the prediction residual data to obtain a signed variable length code; select the one with the smallest number of bits from the unsigned fixed length code, signed fixed length code, unsigned variable length code and signed variable length code, as compressed data. Compared with the existing technology, this application can reduce the bandwidth requirements and power consumption of the chip on the basis of achieving the same image compression quality, and can improve the image compression efficiency and power consumption on the basis of the same chip bandwidth requirements and power consumption. Image compression quality.

Description

Entropy coding method, entropy decoding method and related devices

Technical field

The present application relates to the field of artificial intelligence technology, and in particular to an entropy coding method, an entropy decoding method and related devices.

Background technique

Image compression usually refers to frame compression and video interface compression respectively. Reference frame compression is a type of shallow compression technology that can compress data stored in frame memory to reduce external bandwidth and power consumption. In video compression with an upper bound on the target bit rate, the more reference frames, the higher the prediction accuracy and the better the image quality. However, the main frequency of the chip is generally around 400 MHz. When processing some image data, only A reference frame has exceeded the upper bound of memory processing capabilities, and in practical applications, generally only 70% of the effective bandwidth of the memory working period can be planned, and the quality of video compression is even worse. Therefore, current image compression has high requirements on the bandwidth and power consumption of the chip, which increases the complexity of the chip, thereby reducing the image compression efficiency and reducing the image compression quality.

Contents of the invention

The main purpose of the embodiments of this application is to propose an entropy coding method, an entropy decoding method and related devices, aiming to improve the efficiency and quality of image compression.

In order to achieve the above object, the first aspect of the embodiment of the present application proposes an entropy coding method. The entropy coding method includes:

Obtain only the quantized data and prediction residual data of the frame image data in the original image;

Perform unsigned fixed-length encoding on the quantized-only data to obtain an unsigned fixed-length code;

Determine bit indication information and asymmetric range flag bits according to the prediction residual data;

According to the bit indication information and the asymmetric value field flag bit, the transmission value field of the signed fixed-length encoding is obtained to determine the encoding mode of the signed fixed-length encoding. The asymmetric value field flag bit is in the same The opening and closing conditions of the two corresponding transmission value fields under the bit indication information are opposite;

Perform signed fixed-length coding on the prediction residual data according to the coding method to obtain a signed fixed-length code;

Perform unsigned variable length coding on the quantized-only data to obtain an unsigned variable length code;

Perform signed variable length coding on the prediction residual data to obtain a signed variable length code;

The one with the smallest number of bits is selected from the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code and the signed variable-length code as the compressed data of the frame image data.

In some embodiments, determining bit indication information and asymmetric range flags based on the prediction residual data includes:

According to the prediction residual data, determine the maximum absolute value data of the prediction residual data;

The bit indication information and the asymmetric value range flag are determined according to the absolute maximum data and the corresponding sign of the prediction residual data.

In some embodiments, obtaining a signed fixed-length encoded transmission value domain based on the bit indication information and the asymmetric value domain flag bit includes:

When the bit indication information is greater than 0, determine the boundary information according to the bit indication information;

When the asymmetric value range flag is 1, set the negative end of the boundary information to an open interval and the positive end of the boundary information to a closed interval to obtain the transmission value of the signed fixed-length code area;

When the asymmetric value domain flag bit is 0, the negative end of the boundary information is set as a closed interval and the positive end of the boundary information is set as an open interval to obtain the transmission value of the signed fixed-length code area.

In some embodiments, determining boundary information based on the bit indication information includes:

Subtract 1 from the bit indication information to obtain the subtraction result;

The first preset value is used as the base and the subtraction result is used as the exponent for exponentiation to obtain a positive boundary;

Calculate the opposite number of the corresponding value of the positive boundary to obtain the negative boundary;

Boundary information is determined based on the positive boundary and the negative boundary.

In some embodiments, performing signed variable length coding on the prediction residual data to obtain a signed variable length code includes:

When the prediction residual data is greater than 0 or less than 0, determine the sign bit according to the prediction residual data;

Perform unsigned variable-length coding on the absolute value of the prediction residual data to obtain a first coding;

According to the first code and the sign bit, a signed variable length code is obtained.

In some embodiments, performing unsigned variable length coding on the absolute value of the prediction residual data to obtain the first coding includes:

Add the absolute value of the prediction residual data to the second preset value;

Perform binary conversion on the addition result to obtain the second code;

Determine a coding prefix according to the number of bits of the second code;

According to the encoding prefix and the second encoding, a first encoding is obtained.

In some embodiments, performing signed variable length coding on the prediction residual data to obtain a signed variable length code further includes:

When the prediction residual data is equal to 0, unsigned variable length coding is performed on the prediction residual data to obtain a signed variable length code.

In some embodiments, obtaining only the quantized data and prediction residual data of the frame image data in the original image includes:

Get the frame image data in the original image;

Quantize the frame image data according to preset quantization parameters to obtain only quantized data;

Perform pixel prediction on the quantized-only data to obtain prediction residual data.

In some embodiments, the one with the smallest number of bits is selected from the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code, and the signed variable-length code as the frame. After compressing the image data, the entropy encoding method also includes:

Input the compressed data to a video cache checker for verification;

When the video cache checker is in an overflow state, increase and adjust the quantization parameter, and use the adjusted quantization parameter as a preset quantization parameter;

According to the quantization parameter, repeatedly obtain the compressed data corresponding to the frame image data until the video buffer checker is in a non-overflow state;

Output the compressed data corresponding to when the video cache checker is in a non-overflow state.

In order to achieve the above purpose, the second aspect of the embodiment of the present application proposes an entropy decoding method. The entropy decoding method includes:

Obtain compressed data, the compressed data being generated according to the entropy coding method described in the first aspect;

Determine a coding type according to the compressed data, and the coding type includes unsigned fixed-length coding, signed fixed-length coding, unsigned variable-length coding, and signed variable-length coding;

Decode the compressed data according to the encoding type to obtain decoded data;

Based on the decoded data, frame image data is determined.

In some embodiments, determining frame image data according to the decoded data includes:

When the decoded data is prediction residual data, determine the prediction method and quantization parameters of the decoded data;

Perform reverse prediction on the decoded data according to the prediction method to obtain only quantized data;

The quantized-only data is inversely quantized according to the quantization parameter to obtain frame image data.

In order to achieve the above object, the fourth aspect of the embodiment of the present application proposes an entropy coding device, including:

An acquisition unit used to acquire only quantized data and prediction residual data of the frame image data in the original image;

The first coding unit is used to perform unsigned fixed-length coding on the quantized-only data to obtain an unsigned fixed-length code;

An information determination unit, configured to determine bit indication information and asymmetric value range flags based on the prediction residual data;

A value domain determination unit, configured to obtain the transmission value domain of the signed fixed-length encoding according to the bit indication information and the asymmetric value domain flag bit, so as to determine the encoding method of the signed fixed-length encoding, the asymmetric value The opening and closing conditions of the two transmission value fields corresponding to the field flag bit under the same bit indication information are opposite. The second coding unit is used to perform signed fixed-length coding on the prediction residual data according to the coding method. , get the signed fixed-length code;

The third coding unit is used to perform unsigned variable length coding on the quantized-only data to obtain an unsigned variable length code;

The fourth coding unit is used to perform signed variable length coding on the prediction residual data to obtain a signed variable length code;

A comparison unit configured to select the one with the smallest number of bits from the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code, and the signed variable-length code as the frame image. Compressed data of data.

In order to achieve the above object, a fourth aspect of the embodiment of the present application proposes an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the above is implemented. The entropy coding method described in the first aspect or the entropy decoding method described in the above second aspect.

In order to achieve the above object, the fifth aspect of the embodiment of the present application proposes a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the above-mentioned first aspect is implemented. The entropy coding method described above or the entropy decoding method described in the second aspect above.

The entropy coding method, entropy decoding method and related devices proposed in this application separately encode the quantized data to obtain unsigned fixed-length codes and unsigned variable-length codes, and encode the prediction residual data to obtain signed fixed-length codes. Long codes and signed variable length codes combine variable length coding and fixed length coding, and select the coding method with the best compression efficiency, that is, the coding method corresponding to the coding result with the smallest number of bits, and the corresponding coding result is used as compressed data. In addition, this application determines the transmission value domain of signed fixed-length coding based on the asymmetric value domain flag bit, and the opening and closing conditions of the two transmission value domains corresponding to different asymmetric value domain flag bits under the same indication information are opposite, and it can It improves the information density of signed fixed-length codes and improves coding efficiency. Compared with the existing technology, on the basis of achieving the same image compression quality, this application can effectively reduce the bandwidth requirements and power consumption of the chip, while maintaining the same Based on the chip bandwidth requirements and power consumption, this application can improve the image compression efficiency and improve the image compression quality.

Description of drawings

Figure 1 is a flow chart of an entropy coding method provided by an embodiment of the present application;

Figure 2 is a flow chart of information determination of the entropy coding method provided by the embodiment of the present application;

Figure 3 is a flow chart for determining the transmission value range of the entropy coding method provided by the embodiment of the present application;

Figure 4 is a flow chart for calculating boundary information of the entropy coding method provided by the embodiment of the present application;

Figure 5 is a flow chart of signed variable length code coding of the entropy coding method provided by the embodiment of the present application;

Figure 6 is a flow chart for determining the first encoding of the entropy encoding method provided by the embodiment of the present application;

Figure 7 is another flow chart of signed variable length code coding of the entropy coding method provided by the embodiment of the present application;

Figure 8 is another flowchart of calculating only quantized data and prediction parameters in the entropy coding method provided by the embodiment of the present application;

Figure 9 is a flow chart of the verification of the entropy coding method provided by the embodiment of the present application;

Figure 10 is a main flow chart of the entropy decoding method provided by the embodiment of the present application;

Figure 11 is a flow chart for determining frame image data in the entropy decoding method provided by the embodiment of the present application;

Figure 12 is a schematic diagram of the entropy coding method provided by the embodiment of the present application;

Figure 13 is a schematic diagram of the entropy decoding method provided by the embodiment of the present application;

Figure 14 is a schematic structural diagram of an entropy coding device provided by an embodiment of the present application;

Figure 15 is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application and are not used to limit the present application.

It should be noted that although the functional modules are divided in the device schematic diagram and the logical sequence is shown in the flow chart, in some cases, the modules can be divided into different modules in the device or the order in the flow chart can be executed. The steps shown or described. The terms "first", "second", etc. in the description, claims, and above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific sequence or sequence.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

First, let’s analyze some terms involved in this application:

Artificial intelligence (AI): It is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence; artificial intelligence is a branch of computer science, artificial intelligence Intelligence attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a manner similar to human intelligence. Research in this field includes robotics, language recognition, image recognition, natural language processing, and expert systems. Artificial intelligence can simulate the information process of human consciousness and thinking. Artificial intelligence is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

Machine learning is a data analysis technique that allows computers to do what humans and animals are born with: learning from experience. Machine learning algorithms use computational methods to "learn" information directly from data without relying on predetermined equation models. These algorithms can adaptively improve performance as the number of samples available for learning increases. It is the core of artificial intelligence and the fundamental way to make computers intelligent.

Natural language processing (NLP): NLP uses computers to process, understand and use human languages (such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary subject of computer science and linguistics. It's called computational linguistics. Natural language processing includes syntax analysis, semantic analysis, text understanding, etc. Natural language processing is often used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, and opinion mining. It involves language processing-related Data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research and linguistic research related to language computing, etc.

Information Extraction: Text processing technology that extracts specified types of factual information such as entities, relationships, events, etc. from natural language text and forms structured data output. Information extraction is a technique for extracting specific information from text data. Text data is composed of some specific units, such as sentences, paragraphs, and chapters. Text information is composed of some small specific units, such as words, words, phrases, sentences, paragraphs, or a combination of these specific units. . Extracting noun phrases, person names, place names, etc. from text data is text information extraction. Of course, the information extracted by text information extraction technology can be various types of information.

Video compression, also known as video encoding, refers to the method of converting original video format files into another video format file through compression technology. Video image data has strong correlation, which means there is a lot of redundant information. The redundant information can be divided into air domain redundant information and time domain redundant information. The video compression family is to remove redundant information in the data (removing the correlation between data). Compression technology includes intra-frame image data compression technology, inter-frame image data compression technology and entropy coding compression technology.

Shallow compression, also known as mezzanine compression, is a video compression level that can effectively reduce video bandwidth while maintaining the overall video quality. The compression ratio is usually 2:1 to 8:1.

Reference frame compression is one of the video compression technologies and is a shallow compression technology. Reference frame compression is an effective solution to reduce the memory access bandwidth of the video encoder. It can not only reduce the bandwidth of reading reference frame pixels, but also reduce the bandwidth of the encoder writing reconstructed frame pixels to external memory. Reference frame compression is image compression. Ratio, that is, the ratio of the space occupied before compression to the actual space occupied, is a compression technology that is close to transparent quality in the vicinity of 2 to 3 times.

Entropy coding refers to coding that does not lose any information according to the entropy principle during the coding process. According to the principles of information theory, the optimal data compression encoding method can be found. The theoretical limit of data compression is information entropy. If the amount of information is not lost during the encoding process, that is, information entropy is required to be preserved. This information-preserving encoding is called entropy encoding. It is performed based on the distribution characteristics of the message appearance probability and is a lossless data compression encoding. Common entropy codes include: Shannon coding, Huffman coding and arithmetic coding. In video coding, entropy coding converts a series of element symbols used to represent a video sequence into a compressed code stream for transmission or storage. The input symbols may include quantized transform coefficients, motion vectors, header information (macroblock headers, image headers, sequence headers, etc.) and additional information (mark bit information important for correct decoding).

The processes of entropy decoding and entropy coding correspond one to one. Entropy decoding can be divided into four parts: "judging syntax elements", "arithmetic decoding", "updating context" and "anti-binarization". The "Judging Grammar Elements" part performs simultaneous analysis according to the coding order strictly defined in the standard to obtain which grammar element currently needs to be decoded. The "arithmetic decoding" part first obtains the context model by looking up the table according to the context model type given by the judgment syntax element, and parses the current decoding interval range and offset according to the data decoding engine in the context model, and further obtains the specific decoded binary characters. "Update context" According to the decoded binary characters, the context model is adaptively updated for the next decoding. The "anti-binarization" module restores the binary characters to the real values of the syntax elements, and the parsing of one syntax element is completed. This grammar element will be sent back to the "Judge Grammar Element" module to determine the type of next grammar element and the type of context model.

Image compression usually refers to frame compression and video interface compression respectively. Reference frame compression is a type of shallow compression technology that can compress data stored in frame memory to reduce external bandwidth and power consumption. In video compression with an upper bound on the target bit rate, the more reference frames, the higher the prediction accuracy and the better the image quality. However, the main frequency of the chip is generally around 400 MHz. When processing some image data, only A reference frame has exceeded the upper bound of memory processing capabilities, and in practical applications, generally only 70% of the effective bandwidth of the memory working period can be planned, and the quality of video compression is even worse. Therefore, current image compression has high requirements on the bandwidth and power consumption of the chip, which increases the complexity of the chip, thereby reducing the image compression efficiency and reducing the image compression quality.

Based on this, the present application provides an entropy coding method, an entropy decoding method and related devices. The entropy coding method provided by the embodiment of the present application combines variable length coding and fixed length coding, and selects the coding method with the optimal compression efficiency from them. The encoding result is used as compressed data. In addition, this application determines the transmission value range of the signed fixed-length code based on the asymmetric value range flag bit, which can improve the information density of the signed fixed-length code and improve the coding efficiency. Compared with the existing technology , on the basis of achieving the same image compression quality, this application can effectively reduce the bandwidth requirements and power consumption of the chip, and on the basis of the same chip bandwidth requirements and power consumption, this application can improve the image compression efficiency and improve improves the image compression quality.

The entropy coding method, entropy decoding method and related devices provided by the embodiments of the present application are specifically described through the following embodiments. First, the entropy coding method in the embodiment of the present application is described.

The embodiments of this application can obtain and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or digital computer-controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .

Basic artificial intelligence technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometric technology, speech processing technology, natural language processing technology, and machine learning/deep learning.

The entropy coding method provided by the embodiment of the present application relates to the field of artificial intelligence technology. The entropy coding method provided by the embodiments of the present application can be applied in the terminal or the server, or can be software running in the terminal or the server. In some embodiments, the terminal can be a smartphone, a tablet, a laptop, a desktop computer, etc.; the server can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers. A cloud that can be configured to provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Server; the software can implement the application of entropy coding method, etc., but is not limited to the above form.

The application may be used in a variety of general or special purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, including Distributed computing environment for any of the above systems or devices, etc. The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including storage devices.

It should be noted that in each specific implementation of the present application, when it comes to relevant processing based on user information, user behavior data, user historical data, user location information and other data related to user identity or characteristics, the first step is to perform relevant processing. Obtain the user's permission or consent, and the collection, use and processing of this data will comply with relevant laws, regulations and standards. In addition, when the embodiment of this application needs to obtain the user's sensitive personal information, it will obtain the user's separate permission or separate consent through a pop-up window or jump to a confirmation page. After clearly obtaining the user's separate permission or separate consent, it will then Obtain necessary user-related data for normal operation of the embodiment of the present application.

Referring to Figure 1, Figure 1 is a flow chart of an entropy coding method provided by an embodiment of the present application. The entropy coding method provided by the embodiment of this application includes but is not limited to the following steps:

Step S100: Obtain only quantized data and prediction residual data of the frame image data in the original image.

Step S200: Perform unsigned fixed-length coding on the quantized data to obtain an unsigned fixed-length code.

Step S300: Determine the bit indication information and the asymmetric range flag bit according to the prediction residual data.

Step S400: According to the bit indication information and the asymmetric value domain flag bit, obtain the transmission value domain of the signed fixed-length encoding to determine the encoding method of the signed fixed-length encoding. The asymmetric value domain flag bit corresponds to the same bit indication information. The opening and closing conditions of the two transmission value fields are opposite.

Step S500: Perform signed fixed-length coding on the prediction residual data according to the coding method of signed fixed-length coding to obtain a signed fixed-length code.

Step S600: Perform unsigned variable length coding on the quantized data to obtain an unsigned variable length code.

Step S700: Perform signed variable length coding on the prediction residual data to obtain a signed variable length code.

Step S800: Select the one with the smallest number of bits from the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code, and the signed variable-length code as the compressed data of the frame image data.

It should be noted that the entropy coding method provided by the embodiment of the present application separately codes the quantized data to obtain an unsigned fixed-length code and an unsigned variable-length code, and codes the prediction residual data to obtain a signed fixed-length code. and signed variable-length code, which combines variable-length coding and fixed-length coding, and selects the coding method with the best compression efficiency, that is, the coding method corresponding to the coding result that occupies the smallest code bits, and the corresponding coding result is used as compressed data. In addition , this application determines the transmission value domain of signed fixed-length coding based on the asymmetric value domain flag bit. The opening and closing conditions of the two transmission value domains corresponding to different asymmetric value domain flag bits under the same indication information are opposite, which can improve the efficiency of the code. The information density of the symbol fixed-length code improves the coding efficiency. Compared with the existing technology, on the basis of achieving the same image compression quality, this application can effectively reduce the bandwidth requirements and power consumption of the chip, while using the same chip On the basis of bandwidth requirements and power consumption, this application can improve image compression efficiency and improve image compression quality.

It should be noted that entropy coding is a technology for lossless data compression, which is a common method among image compression methods.

In step S100 of some embodiments, the quantized-only data and the prediction residual data correspond to the frame image data one-to-one respectively. The quantized-only data is generated by quantization of the frame image data and is unsigned data, while the prediction residual data is only Quantized data is predicted by pixels and is a signed number.

In step S200 of some embodiments, the length of the unsigned fixed-length code is fixed, and its length, that is, the bit indication information corresponding to the unsigned fixed-length code is determined by the numerical maximum value of only the quantized data, and the binary code corresponding to the maximum value The most significant bit in is 1, and the higher bits are all 0. The total number of bits starting from the least significant bit of the maximum value to the bit with the most significant bit being 1 is the bits of the unsigned fixed-length code bit length. The length of the unsigned fixed-length code is fixed. No matter how small the value corresponding to the quantized data is, its length is always fixed.

It should be noted that this application first processes unsigned numbers through the minimum unsigned bit method MinUnsignedBit (the maximum value m among the unsigned numbers counted in the quantized data corresponding to the image macroblock). When the unsigned number is input When it is 0, the minimum unsigned number method returns 0, and the minimum unsigned number m is 0. When the input unsigned number is not equal to 0, the reverse bit scanning method BitScanReverse treats the unsigned number as a non-zero unsigned number. Input, return the bit mapping value The minimum number of digits is the bit indication information corresponding to the unsigned fixed-length code. Each unsigned number only takes m bits of the bit indication information of the low-order part of its binary coded number of bits for transmission. In particular, when the bit indication information is equal to 0, there is no need to transmit any unsigned prediction residual data. , because they are both 0 at the same time, only the bit indication information needs to be transmitted.

In step S300 of some embodiments, the bit indication information is the number of bits of the signed fixed-length code, and the asymmetric value field flag bit corresponds to the left and right opening and closing of the transmission value field. The combination of the two can determine the signed fixed-length code. Encoded transfer value range.

In step S400 of some embodiments, the bit indication information is the number of bits of the signed fixed-length code, and the asymmetric value field flag bit corresponds to the left and right opening and closing of the transmission value field. According to this correspondence, the encoding method can be determined, When the bit indication information is the same, the opening and closing conditions of the two transmission value fields corresponding to the asymmetric value field flag bit are opposite. For example, when the bit indication information is 2, the two transmission value fields corresponding to the asymmetric value field flag bit are respectively are (-2,2] and [-2,2).

In step S500 of some embodiments, the signed fixed-length code is set with a sign bit, and the sign bit is used to represent the sign of the prediction residual data.

In step S600 in some embodiments, the length of the unsigned variable length code is a changing value. The length of the unsigned variable length code will change with the size of the prediction residual data. Only the larger the quantized data, the longer the unsigned variable length code. long.

In step S700 of some embodiments, the unsigned fixed-length code is set with a sign bit, and the sign bit is used to represent the sign of the prediction residual data.

In step S800 of some embodiments, the number of bits of unsigned fixed-length codes, unsigned variable-length codes, signed fixed-length codes and signed variable-length codes corresponding to different encoded data is different. , signed fixed-length code, unsigned variable-length code and signed variable-length code, select the one with the smallest number of bits as the compressed data of the frame image data. It occupies the smallest number of code bits, and the corresponding compression efficiency is higher and the compression quality is better. good.

It can be understood that, referring to Figure 2, step S300 includes but is not limited to the following steps:

Step S310: Determine the maximum absolute value of the prediction residual data based on the prediction residual data.

Step S320: Determine the bit indication information and the asymmetric range flag bit according to the sign of the maximum absolute value data and the corresponding prediction residual data.

It should be noted that the bit indication information is the number of bits of the fixed-length code. The block data header corresponding to the fixed-length code usually includes the length indication value of the fixed-length code, that is, the bit indication information. Therefore, the bit indication information is 0 or a positive integer.

It should be noted that through the minimum signed number method MinSignedBit (the maximum absolute value M of the signed number counted in the prediction parameter data corresponding to the image macroblock residual, M corresponds to the positive signed number in the prediction residual data Negative sign case) Obtain the minimum number of digits m of the signed binary number, m is the bit indication information of the signed number. Specifically, in MinSignedBit, MinUnsignedBit is used to find the minimum number of digits n of M; when n is equal to 0, n is returned ; When n is not equal to 0, if the prediction residual data corresponding to M has only positive numbers or only negative numbers, and M is equal to the n-1 power of 2, that is, 2 ^n-1 (that is, 1 is shifted left by n-1 binary bit), return n; when n is not equal to 0, if the prediction residual data corresponding to M has only positive numbers or only negative numbers, and M is not equal to the n-1 power of 2, that is, 2 ^n-1 (that is, 1 When n-1 binary bits are shifted to the left), n+1 is returned; when n is not equal to 0, if the prediction residual data corresponding to M has both positive and negative numbers, n+1 is returned. The asymmetric value range flag bit corresponds to the left and right opening and closing of the transmission value domain in the half-open and half-closed interval. The calculation method is as follows. The minimum number of digits n in M is found through MinUnsignedBit (M). When n is equal to 0, all The prediction residual data are all equal to 0, and there is no need to transmit the asymmetric value domain flag bit, only the bit indication information needs to be transmitted; when n is not equal to 0, if the prediction residual data corresponding to M has only positive numbers and M is equal to 2 ^{n -1} When , the asymmetric value range flag bit is 1; when n is not equal to 0, if the prediction residual data corresponding to M has only negative numbers, and M is equal to 2 ^n-1 , the asymmetric value range flag bit is 0 ; When n is not equal to 0, if the prediction residual data corresponding to M has both positive and negative numbers, the asymmetric value range flag value is 0.

It can be understood that, referring to Figure 3, step S400 includes but is not limited to the following steps:

Step S410: When the bit indication information is greater than 0, determine the boundary information according to the bit indication information.

Step S420: When the asymmetric value range flag is 1, set the negative end of the boundary information to an open interval and the positive end of the boundary information to a closed interval to obtain the transmission value range of the signed fixed-length code.

Step S430: When the asymmetric value range flag is 0, set the negative end of the boundary information to a closed interval and the positive end of the boundary information to an open interval to obtain the transmission value range of the signed fixed-length code.

It should be noted that signed fixed-length codes in the prior art generally have symmetric transmission value fields. Because they are signed, the minimum expression must be expressed by more than two bits, that is, the bit indication information is greater than or equal to 2, for example, 2 When using 2 bits, the value range expressed is a closed interval [-1, 1], which only expresses 3 values. When using 2 bits, 4 values can actually be expressed; when using 3 bits, the value range expressed is [-3, 3], Only 7 values are expressed, and 8 values can actually be expressed with 3 bits;...; the value range expressed with N bits is a closed interval [-2 ^N-1 -1), ((2 ^N-1 -1) ], only expresses 2 ^N -1 values, and with N bits, it can actually express 2 ^N values. Obviously, one value in the signed fixed-length code is a waste. To solve this problem, in the existing technology The transmission value range is a half-open and half-closed interval with the negative end as the closed interval and the positive end as the open interval. The specific range is as follows:

1 bit, expression [-1,1), the value is -1, 0.

2 digits, expressed as [-2,2), the values are -2, -1, 0, 1.

3 digits, expressed as [-4,4), the values are -4, -3, -2, -1, 0, 1, 2, 3.

By analogy, N bits express [-2 ^N-1 ,2 ^N-1 ), and there are 2 ^N values.

The above-mentioned transmission value range is not flexible enough to satisfy the situation where the data in the prediction residual data happens to be in the inverse transmission value range.

It should be noted that for signed fixed-length codes, the block data header of this application adds an asymmetric value domain flag bit to calibrate the left and right opening and closing spaces of the signed fixed-length code value domain. When the asymmetric value domain flag bit is 0 , indicating that the signed fixed-length code adopts the conventional complement truncation method, that is, the negative end of the transmission value range of the signed fixed-length code is a closed interval and the positive end is an open interval. When the asymmetric value range flag is 1, It means that the signed fixed-length code adopts the inverse and then truncated method, that is, the negative end of the transmission value range of the signed fixed-length code is an open interval and the positive end is a closed interval. The setting method of the signed fixed-length code in this application can satisfy the situation that the prediction residual data happens to be in the inverted transmission value domain, improves the flexibility of the signed fixed-length code, and solves the problem of fixed-length code with symmetrical transmission value domain. While solving the problem of code value waste in long codes, it can also improve the information density of asymmetric fixed-length codes.

In step S410 of some embodiments, when the bit indication information is equal to 0, there is no need to calculate or transmit the asymmetric range flag bit, and there is no need to transmit any signed prediction residual data because they are all 0 at the same time. ,Only the bit indication information needs to be transmitted, and when the bit indication information is greater than 0, the boundary information is determined based on the bit indication information.

In step S420 of some embodiments, the asymmetric value domain flag bit corresponds to the left and right opening and closing of the transmission value domain of the fixed-length code. When the bit indication information is greater than 0 and the asymmetric value domain flag bit is 1, the signed fixed-length The negative end of the transmission value range of the long code is an open interval, and the positive end is a closed interval. The boundary information of the transmission value field is determined by the bit indication information. The transmission value range in which the negative end is an open interval and the positive end is a closed interval can adapt to the situation where the positive end value domain dominates.

In step S430 of some embodiments, when the bit indication information is greater than 0 and the asymmetric value domain flag bit is 0, the signed fixed-length code adopts the conventional complement truncation method, that is, the negative number of the transmission value domain of the signed fixed-length code The end is a closed interval, and the positive end is an open interval. The setting method of two fixed-length codes with different transmission value ranges can simultaneously satisfy the situation where the positive end value domain is dominant or the negative end value domain is dominant, thereby improving coding efficiency.

It can be understood that, referring to Figure 4, step S410 includes but is not limited to the following steps:

Step S411: Subtract 1 from the bit indication information to obtain the subtraction result.

Step S412: Use the first preset value as the base and the subtraction result as the exponent for exponentiation to obtain a positive boundary.

Step S413: Calculate the inverse of the value corresponding to the positive boundary to obtain the negative boundary.

Step S414: Determine boundary information based on the positive boundary and negative boundary.

It should be noted that when the bit indication information is N, the exponent is N-1. In addition, the entropy encoding is binary, the first preset value is set to 2, that is, the base is 2, and the result of the exponentiation operation is 2 ^N-1 , that is, the positive boundary is 2 ^N-1 . Calculate the opposite number of the corresponding value of the positive boundary to obtain the negative boundary, then the negative boundary is -2 ^N-1 . The final boundaries include positive boundaries and negative boundaries, and the transmission value range of the signed fixed-length code is determined based on this.

It should be noted that the transmission value range of the signed fixed-length code in step S340 is (-2 ^{N-1,2 N-1} ], and the transmission value range of the signed fixed-length code in step S360 is [-2 ^N-1 ,2 ^N-1 ), the opening and closing conditions of the transmission value fields of the two half-open and half-closed intervals are opposite to each other.

It should be noted that when the asymmetric value domain flag bit is 1, the negative end of the transmission value domain of the signed fixed-length code is an open interval and the positive end is a closed interval. The specific range is as follows:

1 bit, expressing (-1,1], with values 0 and 1.

2 digits, expressing (-2,2], the values are -1, 0, 1, 2.

3 digits, expressing (-4,4], the values are -3, -2, -1, 0, 1, 2, 3, 4.

By analogy, N bits express (-2 ^N-1 ,2 ^N-1 ], and there are 2 ^N kinds of values.

In some embodiments, the preferred encoding scheme is one. The length of the fixed-length code is a fixed value N greater than 0. When the asymmetric range flag is 0, when the prediction residual data is processed, the prediction residual data will be The binary number only takes the low N-bit data indicated by the bit indication information N as the coded value of the prediction residual data. For example, when the bit indication information is 4, -2 can be expressed as 1110, -8 can be expressed as 1000, 0 can be expressed as 0000, and 3 can be expressed as 0011. In addition, when the asymmetric range flag is 1, when processing the prediction residual data, the prediction residual data is inverted and converted into a binary number, and only the low N-bit data indicated by the bit indication information N is taken as the prediction The encoded value of the residual data. For example, when the bit indication information is 4, -2 can be expressed as 0010, -7 can be expressed as 0111, 0 can be expressed as 0000, 8 can be expressed as 1000, and 3 can be expressed as 1101.

In some embodiments, optional encoding end implementation 2, the length of the fixed-length code is a fixed value N greater than 0. When the asymmetric value domain flag is 0, when processing the prediction residual data, the prediction The value corresponding to the residual data is subtracted from the minimum value -2 ^N-1 in the transmission value domain to get the difference, and then the difference is converted into a binary number. Only the low N-bit data indicated by the bit indication information N is taken as the predicted residual data. Encoded value. For example, when the bit indication information is 4, the minimum value -2 ^N-1 in the transmission value field is equal to -8, -2 can be represented as 0110, -8 can be represented as 0000, 0 can be represented as 1000, and 3 can be represented as 1011. In addition, when the asymmetric value domain flag bit is 1, when processing the prediction residual data, the minimum value in the transmission value domain -2 ^N-1 +1 is subtracted from the value corresponding to the prediction residual data to obtain the difference. The difference is then converted into a binary number, and only the lower N-bit data indicated by the bit indication information N is taken as the coded value of the prediction residual data. For example, when the bit indication information is 4 and the minimum value in the transmission value domain -2 ^N-1 +1 is equal to -7, -2 can be represented as 0101, -7 can be represented as 0000, 0 can be represented as 0111, and 8 can be represented is 1111,3 can be expressed as 1010.

The above is just a list of two specific implementations that are relatively simple in terms of algorithm, but does not limit the specific details of the encoding implementation. For example, the number subtracted in the optional encoding implementation two may actually be other constants. But it does not affect the asymmetric value range, which is the real core solution to improve the information density of entropy coding.

It can be understood that, referring to Figure 5, step S700 includes but is not limited to the following steps:

Step S710: When the prediction residual data is greater than 0 or less than 0, determine the sign bit based on the prediction residual data.

Step S720: Perform unsigned coding on the absolute value of the prediction residual data to obtain the first coding.

Step S730: Obtain a signed variable length code based on the first code and the sign bit.

It should be noted that the signed variable-length code of the present application, that is, the setting of the sign bit in the signed fixed-length code more fully saves the effective exponent code, saves the exponent code expression bit in front of the exponent code, and further improves the coding efficiency. .

In step S710 of some embodiments, the sign bit corresponds to the sign of the prediction residual data. When the prediction residual data modulus is a positive integer, the sign bit can be set to 1, and when the sign bit is a negative integer, The sign bit is set to 0. Alternatively, the sign bit can be set to 0 when the prediction residual is a positive integer, and to 1 when the sign bit is a negative integer.

In step S730 of some embodiments, the sign bit is combined with the first code to obtain a signed variable length code. The sign bit can be set before or after the first code. In this application, the sign bit is set after the first code. , is the last position of the signed fixed-length code.

It should be noted that variable length codes are also called K-order variable-length exponential codes and K-order Golomb codes. The Columbus code was originally designed to encode continuous unsigned numbers, but it can be easily mapped and extended to the number field of signed numbers by overlapping positive and negative numbers one by one. The numbers overlap into a sequence: 0, - 1, 1, -2, 2, -3, 3, -4, 4... Give non-negative integers even numbers, give negative integers odd numbers, set x to be a signed number, and index to be the mapped unsigned subscript ,but:

If(x>=0)

index=2x;

Else

index=-2x-1;

It can be seen from the above that in the signed number extension of the variable-length exponent encoding in the prior art, the sign bit directly occupies a code value subscript of the 0-order variable-length exponent encoding to allude to the expression. Specifically, the unsigned 0-order variable-length exponent is expressed The code and the corresponding transmission value set and the signed 0th-order exponential variable length code and the corresponding transmission value set are shown in Table 1.

Table 1

Among them, Variable length exponential coding is used to extend the sign, that is, the sign bit S is set at the end, as shown in Table 2.

Table 2

Compare the number of bits used in the original signed 0th-order exponential variable-length code corresponding to the specific value with the signed 0th-order exponential variable-length code of this application, and the results obtained are shown in Table 3.

table 3

Table 3 only counts the symbols that are very close to the center of the normal distribution. When very close to the center, the probability change is relatively small, but this does not affect the generality of the coding efficiency analysis. The improved method has a greater probability of expressing a given value with fewer bits. The cost is just one extra bit for a small number of values.

It can be understood that, referring to Figure 6, step S720 includes but is not limited to the following steps:

Step S721: Add the absolute value of the prediction residual data to the second preset value.

Step S722: Perform binary conversion on the addition result to obtain the second code.

Step S723: Determine the encoding prefix according to the number of bits of the second encoding.

Step S724: Obtain the first code based on the coding prefix and the second code.

In step S721 of some embodiments, the application is binary encoding, so the second preset value is set to 1.

In step S723 of some embodiments, the encoding prefix corresponds to the number of bits of the second encoding. Specifically, the number of bits of the second encoding is subtracted by 1 to obtain the number of bits of the encoding prefix, and the values of the encoding prefix are all 0.

In step S724 of some embodiments, the encoding prefix is set in front of the second encoding, so that the first encoding can be obtained.

It should be noted that, assuming that the prediction residual data is -7, first add the absolute value of the prediction residual data to the second preset value to get 8, and then perform binary conversion, and the second code obtained is 1000, and the second The number of encoding bits is 4, so the encoding prefix is 000, so the first encoding is 0001000. The prediction residual data is a negative number, and its corresponding sign bit is set to 0. Then the final signed variable length code corresponding to the prediction residual data is 00010000. Alternatively, the prediction residual data is a negative number and its corresponding sign bit is set to 1, then the final signed variable length code corresponding to the prediction residual data is 00010001. When the prediction residual data is positive, the sign bit is processed exactly opposite to when the prediction residual data is negative.

It can be understood that, referring to Figure 7, step S700 also includes but is not limited to the following steps:

Step S740: When the prediction residual data is equal to 0, unsigned variable length coding is performed on the prediction residual data to obtain a signed variable length code.

It should be noted that because 0 is an unsigned number, it is directly encoded with the 0-order unsigned variable-length exponent, and its corresponding signed variable-length code is 1.

It should be noted that in this application, the fixed-length code is used for encoding when the prediction residual data is at a small value. With the ten-digit pixel width shown in Table 4 and the 8x8 data block distance with a width and height of 8, After pixel prediction of this data block, the optimal prediction residual data obtained is 4, the length of the corresponding signed variable length code is 6, and the optimal data byte size is 8*8*6/8=48 bytes , which is better than using a 10-bit unsigned fixed-length code without digital compression. The data byte size is 8*8*10/8=80 bytes. In addition, the maximum value in this data block is 12, and the corresponding binary code bit is 1100. The amount of data corresponding to direct transmission of unsigned encoding is 8*8*4/8=32 bytes. Obviously, the amount of data transmitted through fixed-length code More, better than unsigned encoding and variable length code.

In addition, according to the unsigned fixed-length encoding method described in step S300 of this application, the maximum value in the data block is 12, and the corresponding binary code bit is 1100. According to step S800 of this application, the data bytes corresponding to the unsigned fixed-length encoding are directly transmitted. It is 8*8*4/8=32 bytes. Obviously, less data is transmitted through unsigned fixed-length code, and the information density is higher, which is better than signed variable-length code. In this application, the unsigned variable length encoding directly uses the 0-order unsigned variable length index encoding. If the unsigned variable length code is used, the byte size of the unsigned variable length code corresponding to 0 is 1, and the byte size of the unsigned variable length code corresponding to 4 The byte size is 5, the byte size of the unsigned variable length code corresponding to 8 and 12 is 7, the average bit size is 5, and the corresponding data byte size is 8*8*5/8=40 bytes, still It is unsigned fixed-length encoding that is better.

When the signed fixed-length code is encoded according to the method described in step S374 of this application, the optimal prediction residual data obtained is 4, the optimal bit indication information is 3, the positive end is dominant, and the asymmetric value The domain flag bit is 1, the transmission value domain is (-4,4], the data byte size is 8*8*3/8=24 bytes, only the price of one more asymmetric value domain flag bit is paid. It is The encoding method with the fewest number of bits and the highest information density is used here.

The practical significance and specific implementation of optimal encoding in this application are clearly pointed out here.

Table 4

It can be understood that, referring to Figure 8, step S100 includes but is not limited to the following steps:

Step S110: Obtain frame image data in the original image.

Step S120: Quantize the frame image data according to preset quantization parameters to obtain quantized data.

Step S130: Perform pixel prediction on the quantized-only data to obtain prediction residual data.

It should be noted that the quantized-only data and the prediction residual data correspond to the frame image data one-to-one respectively. The quantized-only data is generated from the frame image data through quantization, which is unsigned data, while the prediction residual data is the quantized-only data through pixels. The predicted value is a signed number. The settings of only quantized data and prediction residual data facilitate unsigned encoding and signed encoding.

It should be noted that quantization can reduce the encoding length without reducing the visual effect and reduce unnecessary information in visual recovery.

It should be noted that prediction mainly includes adjacent element (such as pixel) prediction, string prediction, block prediction, etc. on quantized data to generate prediction residual data. In image coding, macroblocks are the basic unit of coding. In the YUV420 format, each macroblock includes four 8×8 brightness blocks and two 8×8 chrominance blocks, a total of 6 image blocks. In entropy coding, each point of the image block is a coefficient, and this coefficient represents the prediction residual data obtained by the coding step before entropy coding.

It can be understood that, referring to Figure 9, step S800 also includes but is not limited to the following steps:

Step S910: Input the compressed data to the video cache checker for verification.

Step S920: When the video cache checker is in an overflow state, the quantization parameter is increased and adjusted, and the adjusted quantization parameter is used as a preset quantization parameter.

Step S930: Repeatly obtain the compressed data corresponding to the frame image data according to the quantization parameter until the video buffer checker is in a non-overflow state.

Step S940: Output the compressed data corresponding to when the video cache checker is in a non-overflow state.

It should be noted that the Video Buffering Verifier (VBV) is a decoding buffer model that will not overflow when the input code stream complies with the standards of the Digital Video/Audio Compression Experts Group (MovingPictureExpertsGroup, MPEG). Input the obtained compressed data to the video cache checker for verification. When the video cache checker is in the overflow state, the quantization parameters are increased and adjusted, and the adjusted quantization parameters are used as the preset quantization parameters. According to the quantization parameters, repeat Obtain the compressed data corresponding to the frame image data until the video buffer checker is in a non-overflow state, and output the corresponding compressed data when the frequency buffer checker is in a non-overflow state, so that the final compressed data meets MPEG requirements and improves Compression quality.

It should be noted that the image quality degradation in this application is caused by quantization, and whether quantization causes image quality degradation is determined by the buffer model, that is, the video cache checker, that is, by the code rate control.

It should be noted that the structure of compressed data is:

Prediction mode | Whether to use negative long code | Fixed length code length indication value | QP value | Pixel compression data |... | Pixel compression data.

Then, the specific structure of the macroblock header of compressed data is:

Prediction mode | Whether to use negative long code | Fixed length code length indication value | QP value.

Fixed length code transmission:

Prediction mode | Whether to use negative long code | Fixed length code length indication value | QP value.

Variable length code transmission:

Prediction mode | Is it a negative long code | QP value | The byte length of the variable length code pixel compressed data.

It should be noted that the QP value is a quantization parameter. Different pixel bit widths correspond to different maximum quantization parameters. The calculation formula of the maximum quantization parameter MaxQp is as follows:

If (pixel width is an even number)

{

MaxQp=4*pixel width/2;

}

Else

{

MaxQp=4*(pixel width+1)/2;

}

When the pixel bit width is 10, the maximum quantization parameter is 20, and the field length is 5. When the pixel bit width is 7, the maximum quantization parameter is 16, and the field length is 5.

It should be noted that the prediction mode is 3 bits, of which 0 represents upper prediction, 1 represents left prediction, 2 represents upper left prediction, 3 represents upper right prediction, 4 represents DC median prediction, 5 represents quantization mode only, 6 and 7 represent reserve.

It should be noted that the negative long code occupies 1 bit. When the value is 0, the compressed data is a variable length code. When the prediction mode is quantization only mode, that is, unsigned encoding, the compressed data is an unsigned variable length code. , the others are signed variable length codes. When the value corresponding to the negative long code is 1, the compressed data is a fixed-length code. When the prediction mode is quantization-only mode, that is, unsigned coding, the compressed data is an unsigned fixed-length code, and the other is a signed fixed-length code.

It should be noted that the fixed-length code length indication value consists of two parts, 4 of which are bit indication information, whose specific size is the maximum value of the absolute value of the prediction residual data or only quantized data, and the other bit is an asymmetric value. Domain flag bit, this bit is dynamic and only exists when signed fixed-length code is transmitted. If the bit indication information is 0, it means that the prediction residual data or only quantized data to be transmitted is 0. All prediction residual data or only quantized data does not need to be transmitted, and only the bit indication information needs to be transmitted.

It should be noted that, in addition to the macro block header with a fixed byte length, the pixel compression data of the fixed-length code can be directly calculated, that is, there is no need to transmit the byte length information of the compressed data. At the decoding end, it can be obtained from the current macro The block header is positioned directly to the first byte of the next macroblock header.

It should be noted that the macro block header of the variable length code is also of fixed length, which contains the byte length of the optional variable length code pixel compression data. It can also be directly positioned from the current macro block header to the next macro block header at the decoding end. First byte.

It should be noted that, referring to Figure 12, this application first quantizes the frame image data to obtain only quantized data. The only quantized data can be directly subjected to unsigned variable length encoding. First, an unsigned variable length code is obtained, and the unsigned variable length code is obtained. code, that is, the unsigned variable-length code is output to the intermediate buffer, and the number of bits is compared with the unsigned fixed-length code to determine which one is the best, and the unsigned code with better performance is output. In addition, only the quantized data is Pixel prediction, obtain prediction residual data, perform signed coding on the prediction residual data, obtain signed variable length code and signed fixed length code, and output the signed coding with better performance, and then compare the outputs with Signed encoding and unsigned encoding are used to obtain the encoding with the best performance, that is, compressed data. The encoding corresponding to the compressed data occupies the least number of bits. The compressed data is input to the video cache checker for verification. When the video cache checker is in overflow state, the quantization parameters are increased and adjusted, and the adjusted quantization parameters are used as the preset quantization parameters. According to the quantization parameters, the compressed data corresponding to the frame image data is repeatedly obtained until the video cache checker is in a non-overflow state, and the When the video cache checker is in a non-overflow state, the corresponding compressed data is output, so that the final compressed data meets the transmission requirements and improves the compression quality. Based on the statistical results of pixel prediction, this application combines variable-length and fixed-length entropy coding methods, and uses variable-length exponential coding in the signed variable-length code part. Since 0 is an unsigned number, 0-order unsigned variable-length code is used. Index codes are used to express signed numbers other than 0. Setting the sign bit fully saves the effective exponent code and saves the exponent code expression bit in front of the exponent code, which further improves the coding efficiency. In the case of signed fixed-length The code part is set with asymmetric value domain flag bits to calibrate the left and right opening and closing spaces of the signed fixed-length code transmission value domain, which improves the coding efficiency of the fixed-length code.

In addition, referring to Figure 10, this embodiment of the present application also provides an entropy decoding method. The entropy decoding method includes but is not limited to the following steps:

Step S1010: Obtain compressed data. The compressed data is generated according to the above-mentioned entropy coding method.

Step S1020: Determine the encoding type according to the compressed data. The encoding type includes unsigned fixed-length encoding, signed fixed-length encoding, unsigned variable-length encoding, and signed variable-length encoding.

Step S1030: Decode the compressed data according to the encoding type to obtain decoded data.

Step S1040: Determine frame image data according to the decoded data.

It should be noted that the compressed data of the entropy decoding method provided by this application is one of unsigned fixed-length code, unsigned variable-length code, signed fixed-length code and signed variable-length code. First, according to the compressed data, determine Coding type, the coding type corresponds to unsigned fixed-length code, unsigned variable-length code, signed fixed-length code and signed variable-length code. The compressed data is decoded according to the coding type to obtain decoded data. According to the decoded data, determine This entropy decoding method of frame image data corresponds to the above-mentioned entropy encoding method, which can improve the compression efficiency of video compression and improve the compression quality of the video.

It should be noted that the entropy encoding method and entropy decoding method provided by this application can achieve 2-3 times shallow compression with transparent quality. 22 ultra-high definition sequences are now used for encoding and decoding compression tests. The test results are shown in Table 5. Among them, sequence 22 is random noise. It can be seen that the entropy encoding method and entropy decoding method provided by this application can converge well and maintain high quality for unpredictable random noise, while it is completely lossless for the remaining sequence parts, and Achieving error precision that is indistinguishable to the human eye.

table 5

It should be noted that when using reference frame compression or video interface compression, it is very likely to face multiple compression usage scenarios. When the encoding and decoding method provided by this application is serially encoded and decoded with an algorithm with a compression rate of 20 to 30 times, The results are shown in Table 6, and the results when the encoding and decoding method provided by this application are concatenated with an algorithm with a compression rate of 80 to 120 times are encoded and decoded in series, as shown in Table 7.

Table 6

Table 7

It can be seen from Table 6 and Table 7 that in this application, when other compression codecs are connected in series, the impact on quality or bit rate is minimal and can be basically ignored, achieving a transparent series effect.

It can be understood that, referring to Figure 11, step S1040 includes but is not limited to the following steps:

Step S1041: When the decoded data is prediction residual data, determine the prediction method and quantization parameter of the decoded data.

Step S1042: Back-predict the decoded data according to the prediction method to obtain quantized-only data.

Step S1043: Inverse quantize the quantized data according to the quantization parameters to obtain frame image data.

It should be noted that, corresponding to step S130, the decoded data needs to be inversely predicted using a prediction method to obtain quantized-only data. Corresponding to step S120, the quantized-only data needs to be inversely quantized according to the quantization parameters to finally obtain the original image data.

It should be noted that when the decoded data is only quantized data, inverse quantization is directly performed according to step S1043 to obtain frame image data.

It should be noted that, referring to Figure 13, after receiving the compressed data, it is determined whether the compressed data has been predicted. When the compressed data has been predicted, the prediction method and signed coding method of the compressed data, that is, the coding type, are determined according to the prediction method and The compressed data is restored using the signed encoding method. If the compressed data has not been predicted, the unsigned encoding method of the compressed data, that is, the encoding type, is determined, and the compressed data is restored according to the unsigned encoding method. Finally, the restored data is inversely quantized according to the quantization parameters to obtain frame image data.

In addition, referring to Figure 14, this embodiment of the present application also provides an entropy coding device, including:

The acquisition unit 100 is used to acquire only quantized data and prediction residual data of the frame image data in the original image.

The first encoding unit 200 is used to perform unsigned fixed-length encoding on the quantized data to obtain an unsigned fixed-length code.

The information determining unit 300 is configured to determine the bit indication information and the asymmetric value range flag bit according to the prediction residual data.

The value domain determination unit 400 is used to obtain the transmission value domain of the signed fixed-length encoding according to the bit indication information and the asymmetric value domain flag bit, so as to determine the encoding method of the signed fixed-length code. The asymmetric value domain flag bit is in the same The opening and closing conditions of the two corresponding transmission value fields under the bit indication information are opposite.

The second encoding unit 500 is configured to perform signed fixed-length encoding on the prediction residual data according to the encoding method to obtain a signed fixed-length code.

The third encoding unit 600 is used to perform unsigned variable length encoding on the quantized data to obtain an unsigned variable length code.

The fourth encoding unit 700 is used to select the one with the smallest number of bits from the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code, and the signed variable-length code as the compressed data of the frame image data.

The comparison unit 800 is used to compare the number of bits of the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code and the signed variable-length code, and determine the compressed data of the frame image data.

The specific implementation of the entropy coding device is basically the same as the specific implementation of the above-mentioned entropy coding method, and will not be described again here.

An embodiment of the present application also provides an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the above-mentioned entropy encoding method or entropy decoding method. The electronic device can be any smart terminal including a tablet computer, a vehicle-mounted computer, etc.

Please refer to Figure 15. Figure 15 illustrates the hardware structure of an electronic device according to another embodiment. The electronic device includes:

The processor 901 can be implemented by a general CPU (Central Processing Unit, central processing unit), a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement The technical solutions provided by the embodiments of this application.

The memory 902 can be implemented in the form of read-only memory (ReadOnlyMemory, ROM), static storage device, dynamic storage device, or random access memory (RandomAccessMemory, RAM). The memory 902 can store operating systems and other application programs. When implementing the technical solutions provided by the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 902 and called by the processor 901 to execute the implementation of this application. Examples of entropy encoding methods or entropy decoding methods.

Input/output interface 903 is used to implement information input and output.

The communication interface 904 is used to realize communication interaction between this device and other devices. Communication can be achieved through wired methods (such as USB, network cables, etc.) or wireless methods (such as mobile network, WIFI, Bluetooth, etc.).

Bus 905 carries information between the various components of the device (eg, processor 901, memory 902, input/output interface 903, and communication interface 904).

The processor 901, the memory 902, the input/output interface 903 and the communication interface 904 implement communication connections between each other within the device through the bus 905.

Embodiments of the present application also provide a computer-readable storage medium that stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned entropy encoding method or entropy decoding method.

As a non-transitory computer-readable storage medium, memory can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory may include high-speed random access memory and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory located remotely from the processor, and the remote memory may be connected to the processor via a network. Examples of the above-mentioned networks include but are not limited to the Internet, intranets, local area networks, mobile communication networks and combinations thereof.

The entropy coding method, the entropy decoding method and related devices provided by the embodiments of the present application. The entropy coding method separately codes the quantized data to obtain an unsigned fixed-length code and an unsigned variable-length code, and codes the prediction parameters to obtain Signed fixed-length codes and signed variable-length codes combine variable-length coding and fixed-length coding, and select the coding method with the best compression efficiency, that is, the coding method corresponding to the coding result that occupies the smallest code bits, and the corresponding coding result As compressed data, in addition, this application determines the transmission value domain of the signed fixed-length code based on the asymmetric value domain flag bit, which can improve the information density of the signed fixed-length code and improve the coding efficiency. Compared with the existing technology, On the basis of achieving the same image compression quality, this application can effectively reduce the bandwidth requirements and power consumption of the chip. On the basis of the same chip bandwidth requirements and power consumption, this application can improve the compression efficiency of the image and improve the image quality. compression quality.

The embodiments described in the embodiments of the present application are for the purpose of more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art will know that with the evolution of technology and new technologies, As application scenarios arise, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

Those skilled in the art can understand that the technical solutions shown in the figures do not limit the embodiments of the present application, and may include more or fewer steps than those shown in the figures, or combine certain steps, or different steps.

The device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separate, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Those of ordinary skill in the art can understand that all or some steps, systems, and functional modules/units in the devices disclosed above can be implemented as software, firmware, hardware, and appropriate combinations thereof.

The terms "first", "second", "third", "fourth", etc. (if present) in the description of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe specific objects. Sequence or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances so that the embodiments of the application described herein can be practiced in sequences other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, e.g., a process, method, system, product, or apparatus that encompasses a series of steps or units and need not be limited to those explicitly listed. Those steps or elements may instead include other steps or elements not expressly listed or inherent to the process, method, product or apparatus.

It should be understood that in this application, "at least one (item)" refers to one or more, and "plurality" refers to two or more. "And/or" is used to describe the relationship between associated objects, indicating that there can be three relationships. For example, "A and/or B" can mean: only A exists, only B exists, and A and B exist simultaneously. , where A and B can be singular or plural. The character "/" generally indicates that the related objects are in an "or" relationship. “At least one of the following” or similar expressions thereof refers to any combination of these items, including any combination of a single item (items) or a plurality of items (items). For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c" ”, where a, b, c can be single or multiple.

In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the above units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or may be Integrated into another system, or some features can be ignored, or not implemented. On the other hand, the coupling or direct coupling or communication connection between each other shown or discussed may be through some interfaces, and the indirect coupling or communication connection of the devices or units may be in electrical, mechanical or other forms.

The units described above as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or software functional units.

Integrated units may be stored in a computer-readable storage medium if they are implemented in the form of software functional units and sold or used as independent products. Based on this understanding, the technical solution of the present application is essentially or contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including multiple instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc. that can store programs. medium.

The preferred embodiments of the embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of rights of the embodiments of the present application. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of rights of the embodiments of the present application.

Claims (14)

1. An entropy encoding method, characterized in that the entropy encoding method comprises:

obtaining only quantized data and prediction residual data of frame image data in an original image, wherein the only quantized data is generated after the frame image data is quantized, and the prediction residual data is obtained by carrying out pixel prediction on the only quantized data;

performing unsigned fixed-length coding on the quantized data only to obtain an unsigned fixed-length code;

determining bit indication information and an asymmetric value range flag bit according to the prediction residual data;

obtaining a transmission value range with a symbol fixed-length code according to the bit indication information and the asymmetric value range zone bit to determine a coding mode of the symbol fixed-length code, wherein the asymmetric value range zone bit is opposite to the two corresponding transmission value ranges under the same bit indication information;

Carrying out signed fixed-length coding on the prediction residual data according to the coding mode of the signed fixed-length coding to obtain a signed fixed-length code;

performing unsigned variable length coding on the quantized data only to obtain an unsigned variable length code;

carrying out signed variable length coding on the predicted residual data to obtain a signed variable length code;

and selecting one with the least bit number from the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code and the signed variable-length code as compressed data of the frame image data.

2. The entropy encoding method according to claim 1, wherein the determining bit indication information and an asymmetric value range flag bit according to the prediction residual data comprises:

determining maximum absolute value data of the prediction residual data according to the prediction residual data;

and determining the bit indication information and the asymmetric value range zone bit according to the maximum absolute value data and the corresponding sign of the prediction residual error data.

3. The entropy encoding method according to claim 1, wherein the obtaining a signed fixed length encoded transmission range according to the bit indication information and the asymmetric range flag bit comprises:

Determining boundary information according to the bit indication information when the bit indication information is greater than 0;

when the asymmetric value range flag bit is 1, setting a negative end of the boundary information as an open section and setting a positive end of the boundary information as a closed section to obtain the transmission value range of the signed fixed-length code;

and when the asymmetric value range flag bit is 0, setting the negative end of the boundary information as a closed interval and the positive end of the boundary information as an open interval to obtain the transmission value range of the signed fixed-length code.

4. The entropy encoding method of claim 3, wherein the determining boundary information according to the bit indication information comprises:

subtracting 1 from the bit indication information to obtain a subtraction result;

performing power operation by taking a first preset numerical value as a base and taking the subtraction result as an exponent to obtain a positive boundary;

calculating the opposite number of the corresponding value of the positive boundary to obtain a negative boundary;

and determining boundary information according to the positive boundary and the negative boundary.

5. The entropy encoding method according to claim 1, wherein the performing signed variable length encoding on the prediction residual data to obtain a signed variable length code comprises:

When the predicted residual data is more than 0 or less than 0, determining a sign bit according to the predicted residual data;

performing unsigned variable length coding on the absolute value of the prediction residual data to obtain a first code;

and obtaining a signed variable length code according to the first code and the sign bit.

6. The entropy encoding method according to claim 5, wherein the performing unsigned variable length encoding on the absolute value of the prediction residual data to obtain a first code comprises:

adding the absolute value of the prediction residual data to a second preset value;

binary conversion is carried out on the added result to obtain a second code;

determining a coding prefix according to the second coded bit number;

and obtaining a first code according to the code prefix and the second code.

7. The entropy encoding method of claim 5, wherein the performing signed variable length encoding on the prediction residual data to obtain a signed variable length code, further comprises:

and when the predicted residual data is equal to 0, performing unsigned variable length coding on the predicted residual data to obtain a signed variable length code.

8. The entropy encoding method according to claim 1, wherein the acquiring only quantization data and prediction residual data of frame image data in an original image comprises:

Acquiring frame image data in an original image;

quantizing the frame image data according to preset quantization parameters to obtain quantized data only;

and carrying out pixel prediction on the quantized data only to obtain prediction residual data.

9. The entropy encoding method according to claim 8, wherein after selecting one of the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code, and the signed variable-length code having the smallest number of bits as the compressed data of the frame image data, the entropy encoding method further comprises:

inputting the compressed data to a video cache checker for checking;

when the video cache checker is in an overflow state, increasing and adjusting the quantization parameter, and taking the adjusted quantization parameter as a preset quantization parameter;

repeatedly acquiring compressed data corresponding to the frame image data according to the adjusted quantization parameter until the video buffer checker is in a non-overflow state;

outputting the compressed data corresponding to the video buffer checker in a non-overflow state.

10. An entropy decoding method, comprising:

Acquiring compressed data, the compressed data being generated according to the entropy encoding method of any one of claims 1 to 9;

determining a coding type according to the compressed data, wherein the coding type comprises an unsigned fixed-length code, a signed fixed-length code, an unsigned variable-length code and a signed variable-length code;

decoding the compressed data according to the coding type to obtain decoded data;

and determining frame image data according to the decoded data.

11. The entropy decoding method of claim 10, wherein determining frame image data from the decoded data comprises:

when the decoded data is prediction residual data, determining a prediction mode and a quantization parameter of the decoded data;

performing inverse prediction on the decoded data according to the prediction mode to obtain quantized data only;

and dequantizing the quantized data only according to the quantization parameter to obtain frame image data.

12. An entropy encoding device, comprising:

the acquisition unit is used for acquiring only quantized data and prediction residual data of frame image data in an original image, wherein the only quantized data is generated after the frame image data is quantized, and the prediction residual data is obtained by carrying out pixel prediction on the only quantized data;

The first coding unit is used for carrying out unsigned fixed-length coding on the quantized data only to obtain an unsigned fixed-length code;

the information determining unit is used for determining bit indication information and an asymmetric value range zone bit according to the prediction residual data;

the value range determining unit is used for obtaining a transmission value range with a symbol fixed-length code according to the bit indication information and the asymmetric value range zone bit so as to determine a coding mode of the symbol fixed-length code, wherein the opening and closing conditions of the two corresponding transmission value ranges of the asymmetric value range zone bit under the same bit indication information are opposite;

the second coding unit is used for carrying out signed fixed-length coding on the prediction residual data according to the coding mode of the signed fixed-length coding to obtain a signed fixed-length code;

the third coding unit is used for performing unsigned variable length coding on the quantized data only to obtain an unsigned variable length code;

a fourth coding unit, configured to perform signed variable length coding on the prediction residual data to obtain a signed variable length code;

and the comparison unit is used for selecting one with the least bit number from the unsigned fixed-length code, the signed fixed-length code, the unsigned variable-length code and the signed variable-length code as compressed data of the frame image data.

13. An electronic device comprising a memory storing a computer program and a processor implementing the entropy encoding method of any one of claims 1 to 9 or the entropy decoding method of any one of claims 10 to 11 when the computer program is executed.

14. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the entropy encoding method of any one of claims 1 to 9 or the entropy decoding method of any one of claims 10 to 11.