CN104301725A - A CAVLD Method Based on Index Search - Google Patents

Abstract

Disclosed in the invention is an indexing-based adaptive variable length decoding (CAVLD) searching method. The method comprises the following steps that: S1, according to value ranges of different syntax elements NC, a variable-length code table corresponding to a corresponding element is selected; S2, an input code stream is read, and the number of units with continuous prefixes of 0 in the code stream is calculated and is used as a first level index number; S3, on the basis of an intrinsic relation between the number of the codon prefixes 0 and the code length, a corresponding codon length is found out; S4, if the found codon length has a unique value, the codon suffix length is directly determined according to a formula; and if the codon length has two probable values, one-bit bit stream is read again to determine the code length and then the codon suffix length is determined according to a formula; S5, the bit stream of the codon suffix length is taken out directly from the input code stream and the value is used as a second level index number; and S6, a decoding condon is determined rapidly based on the first level index number and the second level index number. According to the invention, searching is carried out by using the indexing technology; the TLSS internal memory consumption can be reduced; the lookup speed can be accelerated; and the code table storage space can be saved.

Description

A CAVLD Method Based on Index Search

technical field

The invention relates to an index-based search CAVLD method. the

Background technique

H.264/AVC is jointly formulated by ITU-VCEG (International Telecommunications Union Telecommunication) Video Coding Expert Group, ISO/IEC (International Organization for Standardization/International Electrotechnical Commission) International Standardization Organization and International Electrotechnical Commission A relatively advanced international standard for video compression coding. Because of its high-efficiency coding and compression performance, friendly network interface, and the ability to provide high-quality video and image transmission on a small bandwidth, it has been widely used in the fields of mobile video communication and image storage compression. It is the most influential industry standard. CAVLD (Adaptive Variable Length Decoding) is a common encoding entropy decoding method in the H.264/AVC standard, because it can dynamically switch among multiple sets of code tables with different structures according to different video streams, thereby greatly improving video quality. Coding efficiency and quality purposes. However, while CAVLD improves the efficiency and quality of video coding, it consumes a lot of power consumption of the video decoder because of its frequent lookup of VLTs (Variable Length Tables). Therefore, the high performance of H.264/AVC is achieved by at the expense of increasing its decoder power consumption. the

As we all know, frequent table lookup operations consume a lot of memory and lookup time, which in turn produces a large power consumption. However, in the standard H.264/AVC, CAVLD needs to decode five syntax elements: Coeff_token, T1s, Level, Run_before and Total_zeros. Among them, decoding the three syntax elements of Coeff_token, Run_before and Total_zeros needs to frequently search the variable-length code table for decoding, so the code search in CAVLD takes a lot of memory and search time of the H.264/AVC video decoder, which in turn consumes a lot of power consumption. Therefore, how to reduce frequent code table lookup operations in CAVLD, thereby reducing memory consumption and time when searching for variable-length code tables, and saving power consumption of the decoder has become an important issue we are currently concerned about. Aiming at the above problem, some scholars have carried out some research and put forward some valuable solutions. the

In terms of hardware design, Song Na and others proposed to use Verilog HDL technology to realize the CAVLD decoder, and use simple addition operations to realize time-consuming table lookup operations and improve the decoding speed. Heng et al proposed a method to combine all code tables into a code table, and then use the form of sub-code tables to organize and search for this code table; Lee et al. proposed a method of utilizing the pipeline structure to realize conventional operations; Wang et al. designed a new type of decoder with low power consumption and high performance to improve the overall decoding performance of CAVLD; Fang et al proposed to use the combined effect of CMOS and FPGA technology to improve the performance of CAVLD. In the software design part, a general method based on binary tree table lookup is proposed to improve the time performance of CAVLD table lookup, but this method is not very effective in some systems due to some uncertainties in obtaining memory. Li Fen et al. proposed [14] to divide the code table according to the different prefixes of the code words, which narrowed the scope of the table lookup and reduced the traversal of the entire code table to reduce the decoding time of long code words. Moon et al. proposed to use integer arithmetic operators to realize the code table lookup method of Run_before and Total_zeros elements, which can greatly reduce the consumption of memory and time when CAVLC decodes CAVLD table lookup; Lu et al. proposed a new entropy decoding algorithm, which can improve 75.1%-82.7% decode time. Lee et al. proposed a new code table structure and a new look-up method to greatly reduce memory consumption and improve the speed of decoding look-up tables. Kim et al. proposed to use the method based on integer arithmetic operators to realize other element table lookup methods except Coeff_token and Run_before elements, thereby greatly saving the memory consumption of the lookup table and reducing the time of the lookup table. Jingjing Han and others proposed a table lookup process optimization method that divides the CAVLD code table into regions, accurately looks up the table range, and then performs entropy decoding. Uchihara et al. proposed a fast change method for the level coding change problem in CAVLD, which can reduce the energy consumption of CAVLD amplitude change by 70%. For the software CAVLD decoder structure, Uchihara et al. proposed a method based on amplitude The length extraction method, the proposed method can improve the decoder speed by 22%. the

The present invention mainly aims at the problems that TLSS occupies a large amount of memory when looking up the code table in the CAVLD process in the current H.264/AVC standard, takes a long time to look up the table, and consumes a lot of storage space in the code table, and proposes a CAVLD based on index search The method uses the number of codeword prefix 0 as the first-level index number, uses the codeword suffix value as the second-level index number, and finds the decoding codeword by determining the first and second-level index numbers, and realizes the CAVLD method based on the index search, The number of decoding codeword search operations is reduced, memory consumption and search time are saved when searching the code table, and power consumption in the CAVLD is further saved. The invention improves the TLSS method in the standard CAVLD, expands the existing table look-up technology in the CAVLD, greatly improves the decoding performance of the CAVLD, and makes it more efficient and suitable for small area and low power consumption embedded in the mobile environment. The product completes the fast decoding requirements. the

The purpose of the foregoing discussion is to introduce the reader to various aspects of the technology that may be related to various aspects of the present invention that will be described and/or claimed below. It is believed that the discussion will help provide background information for the reader to facilitate a better With the various aspects of the present invention in mind, it is therefore to be understood that these discussions are to be read in this light, and not as admissions of prior art. the

Contents of the invention

The present invention mainly aims at the problem that the TLSS method in the current standard CAVLD consumes a lot of memory, takes a long time to look up the table, and consumes a large storage space of the code table when looking up the table, and proposes a method for searching the CAVLD based on an index, which improves the standard CAVLD process. The existing CAVLD table look-up technology is expanded, and the decoding ability of CAVLD on H.264/AVC data stream is greatly improved. the

The purpose of the present invention is achieved through the following technical solutions:

Provide an index-based search CAVLD method, including the following processing steps:

S1, according to the value range of different syntax elements NC, select the variable length code table corresponding to the corresponding element;

S2. Read the input code stream, calculate the number of continuous prefix 0 in the input code stream and use it as the first-level index number;

S3, according to the inherent relationship between the number of codeword prefix 0 and the code length, find the corresponding codeword length;

S4, if the found code length has a unique value, then according to the formula: code word suffix length=code word length-the number of code word prefix 0, directly determine the code word suffix length; if the code length found by hash table There are 2 possible values for the length, then the code length can be determined after reading the 1-bit bit stream again and judging whether its value is zero; then according to the formula: code word suffix length = code word length - code word prefix 0 number, determine the codeword suffix length;

S5, directly fetch the bit stream of the codeword suffix length from the input code stream, and use its value as the second-level index number;

S6. Quickly determine the decoding codeword according to the first and second level index numbers determined above. the

Wherein, in step S1, in step S1, the Coeff_token element corresponds to three variable-length code tables, and different variable-length code tables are selected according to the values of different NCs. the

Wherein, in step S4, only when two possible values are found in the code word length, it is necessary to read 1 bit stream to judge its value to determine the code length. There is a one-to-one correspondence between the code lengths, and the corresponding code length can be obtained directly from the number of codeword prefix 0s. the

The beneficial effects of the present invention are: the proposed index-based search method for CAVLD adopts index technology to improve the table lookup operation in standard CAVLD, and overcomes the large memory consumption of the TLSS method in the current standard CAVLD when looking up tables. The shortcomings of taking a long time and occupying a large storage space of the code table have greatly improved the performance of the standard CAVLD, making it more effective to meet the fast decoding requirements of portable products with low power consumption and small area in the mobile environment. The invention improves the code table lookup process of the TLSS method in the CAVLD in the current standard, expands the existing CAVLD table lookup technology, and greatly improves the decoding ability of the CAVLD on the H.264 data stream. the

Description of drawings

Utilize accompanying drawing to further illustrate the present invention, but the embodiment in the accompanying drawing does not constitute any restriction to the present invention, for those of ordinary skill in the art, under the premise of not paying creative work, can also obtain other according to following accompanying drawing attached drawings. the

Fig. 1 is the algorithm work flowchart of the present invention. the

Fig. 2 is an example diagram of decoding Coeff_token elements using the algorithm of the present invention. the

FIG. 3 is a comparison diagram of memory consumption between the algorithm of the present invention and the TLSS algorithm when testing different image series (30 frames, QP=24). the

Fig. 4 is a comparison diagram of memory consumption between the algorithm of the present invention and the TLSS algorithm when testing different image series (90 frames, QP=32). the

FIG. 5 is a comparison diagram of memory consumption between the algorithm of the present invention and the TLSS algorithm when testing different QP values (Mobile (CIF), 30 frames). the

FIG. 6 is a comparison diagram of memory consumption between the algorithm of the present invention and the TLSS algorithm at different QP values (Walk (QCIF), 90 frames). the

Detailed ways

In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments. And the features in the embodiments can be combined with each other. the

In this example, the VC++6.0 development tool is used as the system software development and simulation platform, and JM16.2 is selected as the test model. Table 1 shows the setting values of some important environmental parameters in the test model, as follows. the

Table 1

Table 2 is a hash table structure diagram established by the algorithm of the present invention, specifically as follows. the

Table 2

The experiment selects six different format image series (CIF and QCIF) such as Mobile(M), Walk(W), Paris(P), Football(F), Soccer(S) and Walk(W) as the test series. Under conditions and environments, by testing the above-mentioned series under different series of frame lengths, QP values and other parameters, to compare the memory consumption when the method proposed by the present invention and the TLSS standard method look up the code table, time consumption and space consumption when storing the code table and so on, so as to investigate the overall performance of the algorithm proposed by the present invention. the

Figure 1 is a flowchart of the index-based CAVLD method proposed in this paper. Assuming that the decoding input code stream is 0000100011..., and the value of NC is 0≤NC<2, taking the decoding of Coeff_token as an example, we illustrate our proposed index-based search CAVLD method. Specifically, it can be divided into the following steps:

S1. According to the value of NC 0≤NC<2, select VLCT0 of Coeff_token as the program entry;

S2, read the input code stream, calculate the number of continuous prefix 0 of the code word in the input code stream 0000100011 to be 4, and use it as the first-level index number;

S3. According to the corresponding relationship between the number 4 of codeword prefix 0 and the codeword length (as shown in Table 2), the codeword length can be obtained as 6 or 7. the

S4, because there are two possible values 6 or 7 in the codeword length corresponding to the number (4) of the codeword prefix 0, it is necessary to read 1 bit stream operation again, because the bit stream value of reading this bit is 0, Determine the code length corresponding to the number (4) of codeword prefix 0 to be 7, then according to the formula: codeword suffix length=codeword length-codeword The number of prefix 0 determines that the codeword suffix length is 2;

S5, according to the length 2 of codeword suffix, read 2 bit streams from code stream, and use its value (0) as the second level index number;

S6. Search the index table according to the first and second level index numbers determined above. As shown in Table 3, the decoding codeword can be quickly determined to obtain the corresponding decoding codeword 0x65. the

table 3

Fig. 3 is the decoding process of decoding the Coeff_token element using the method of the present invention. Note that in this example, the decoding process of the Run_before and Total_zeros elements is similar to that of the Coeff_token element. the

Fig. 3 and Fig. 4 are comparison diagrams of memory consumption when testing different image series (30 frames, QP=248 and 90 frames, QP=32) between the algorithm of the present invention and the TLSS algorithm respectively. Fig. 5 and Fig. 6 are comparison diagrams of memory consumption when testing different QP values (Mobile (CIF), 30 frames and Walk (QCIF), 90 frames) between the algorithm of the present invention and the TLSS algorithm respectively. It can be seen from Fig. 3 to Fig. 6 that no matter in different sequences of images or under different QP value sequence tests, the method of the present invention is lower than the TLSS method in terms of the memory consumption of the search code table, and can save about 35% of the memory of the code table consume. The main reason for the analysis of the savings is that this method uses the index search method to reduce the codeword suffix length and the number of consecutive search times of the codeword in the TLSS method, thereby saving part of the memory consumption when looking up the code table and improving the performance of CAVLD. the

Table 4 shows that the algorithm of the present invention and the standard TLSS algorithm spend time in code table lookup when decoding different image sequences. the

Table 4

It can be seen from Table 4 that the index-based search method proposed by the present invention is superior to the standard TLSS method in terms of code table search time, and can save about 40% of the code table search time. The main reason for the reduction is that the method proposed in the present invention uses the index search technology to accelerate the determination of the codeword suffix length and the decoding codeword, thereby reducing the codeword suffix length and the decoding codeword search time. the

Table 5 is a comparison of the code table storage space consumption between the algorithm of the present invention and the standard TLSS method. the

table 5

As can be seen from Table 5, compared with the standard TLSS method, our proposed method can save 923 bytes of code table storage space, and the main reason for saving is that the proposed method of the present invention uses an index table to store the codtab() element in the standard TLSS , thereby saving the storage space of the codtab() code table in the TLSS, and improving the utilization rate of the storage space of portable devices with low power consumption and limited storage resources in the mobile environment. the

According to the number of code word prefix 0 calculated from the input code stream, the present invention uses the hash table search technology to find the code word suffix length from the established hash table, which can reduce the search, matching and judgment of the code word suffix length The operation saves the memory consumption and table lookup time of table lookup in CAVLD. After determining the codeword suffix length, according to the inherent relationship of code length = number of codeword prefix 0s + codeword suffix length, a method for quickly determining the code length can be found, which can reduce the operations of searching, matching, judging and determining the code length, and save searching Table memory consumption and table lookup time. Because the hash table technology is used to store the codeword suffix length, the code length can be quickly determined according to the number of codeword prefix 0+codeword suffix length=code length, which can save the lentab() search step in the standard TLSS method and The lentab() code table storage space improves the utilization rate of the code table storage space in CAVLD, and is a relatively efficient CAVLD method. the

Use the inherent relationship between the number of codeword prefix 0 and the codeword length to determine the codeword length, saving the time for searching, matching, judging and determining the codeword length;

According to the inherent relationship between codeword suffix length = codeword length - the number of codeword prefix 0, a method for quickly determining the number of codeword suffix digits and values is found, which saves codeword suffix search, matching, judgment and determination time;

By using the number of codeword prefix 0 as the first-level index number and the codeword suffix value as the second-level index number, the determination of the above-mentioned first and second-level index numbers can quickly determine the decoding codeword, and realize searching CAVLD based on the index method;

It can not only save memory consumption and decoding time when CAVLD looks up the table, but also save code table storage space, which is a relatively efficient CAVLD method. the

Many specific details have been set forth in the above description to facilitate a full understanding of the present invention. However, the present invention can also be implemented in other ways than those described here, therefore, it should not be construed as limiting the protection scope of the present invention. the

In a word, although the present invention has exemplified the above-mentioned preferred embodiments, it should be noted that although those skilled in the art can make various changes and modifications, unless such changes and modifications deviate from the scope of the present invention, they should include Within the protection scope of the present invention. the

Claims (3)

1. A kind of CAVLD method based on index search, it is characterized in that: comprise following processing steps: S1, according to the value range of different syntax elements NC, select the variable length code table corresponding to the corresponding element; S2. Read the input code stream, calculate the number of continuous prefix 0 in the input code stream and use it as the first-level index number; S3, according to the inherent relationship between the number of codeword prefix 0 and the code length, find the corresponding codeword length; S4, if the found code length has a unique value, then according to the formula: code word suffix length=code word length-the number of code word prefix 0, directly determine the code word suffix length; if the code length found by the hash table There are 2 possible values for the length, then the code length can be determined after reading the 1-bit bit stream again and judging whether its value is zero; then according to the formula: code word suffix length = code word length - code word prefix 0 The number determines the codeword suffix length; S5. Directly fetch the bit stream of the codeword suffix length from the input code stream, and use its value as the second-level index number; S6. Quickly determine the decoding codeword according to the first and second level index numbers determined above. 2. The index-based search CAVLD method according to claim 1 is characterized in that: in step S1, in step S1, the Coeff_token element corresponds to three variable-length code tables, and selects different code tables according to the value of its different NCs. Variable length code table. 3. The index-based search CAVLD method according to claim 1 or 2, characterized in that: in step S4, only when two possible values are found in the length of the code word, it is necessary to read 1 bit stream to judge its value To determine the code length, in most cases, since there is a one-to-one correspondence between the number of code word prefix 0 and the code length, the corresponding code length can be obtained directly from the number of code word prefix 0.