CN104717499B - A kind of storage method of huffman table and the Hofmann decoding method for JPEG - Google Patents

Abstract

A kind of storage method the present invention provides huffman table and the Hofmann decoding method for JPEG, in decoding, 4 JPEG code streams are read in every time, using 4 JPEG code streams as address search huffman table, obtain storage data, and the weights of Huffman code word are obtained accordingly, so as to complete JPEG decodings.Middle use is serial by bit decision compared with the prior art, with higher efficiency；Simultaneously, its storage method only by changing huffman table, in Hofmann decoding without using the logical operation of any complexity, simply simply read the memory of huffman table, and pass through effective marker position and go to judge code word, complicated decoding circuit is not required, thus also lessens the complexity of Hofmann decoding.

Description

A kind of storage method of huffman table and the Hofmann decoding method for JPEG

Technical field

The present invention relates to video coding and decoding technology field, the storage method of more particularly to a kind of huffman table and for JPEG Hofmann decoding method.

Background technology

JPEG (Joint Photographic Experts Group) standard is using huffman algorithm as data encoding Algorithm.Huffman (Huffman, also referred to as Huffman) coding is a kind of Run- Length Coding of variable length, and main thought is to higher The data of probability use less bits of encoded, and use longer bits of encoded to the data of low probability, so as to fulfill data The purpose of compression.In JPEG decoding ends, it is common practice to first recover the huffman table in code stream, and store, Ran Hou The code word that Huffman code is carried out in compressed data judges, is searched by code word in huffman table and obtains corresponding weights, so as to Realize Hofmann decoding.

Since the huffman table in JPEG code streams is generated based on image statistics, so in different code streams, suddenly The content and length of Fu Manbiao is different.Therefore, when carrying out code word judgement, traditional way generally using it is serial by Bit decision or parallel comparator judge.These method logics complexity and waste of resource, while cause entire Hough The working frequency of graceful decoding circuit is limited.

Therefore, how to reduce the complexity of Hofmann decoding, improve the efficiency of Hofmann decoding into people in the art Member's problem urgently to be resolved hurrily.

The content of the invention

A kind of storage method it is an object of the invention to provide huffman table and the Hofmann decoding method for JPEG, To solve the problems, such as that existing Hofmann decoding complexity is high, efficiency is low.

In order to solve the above technical problems, the present invention provides a kind of storage method of huffman table, the huffman table is deposited Method for storing includes：Huffman table is stored using a memory space, the memory space is divided into 1 or multiple memory blocks, described Memory space includes the storage unit of multiple equal lengths, wherein, each storage unit is indicated by an address, and each storage is single The length of member is at least 11 bits, in 11 bits, 1 bit storage code word effective marker position, 2 bit storage code words Effective length, the corresponding weights of 8 bit storage code words or offset address；

Wherein, in a storage unit, in 11 bits, when the code word effective marker position of storage represents invalid, The code word effective length of storage is 0, and the offset address for being directed to next memory block of remaining 8 storages；In a storage In unit, in 11 bits, when the code word effective marker position of storage represents effective, the code word effective length of storage can be sentenced Belong to the bit of current value in disconnected current code word, the corresponding weights of 8 bit storage code words can simultaneously be read according to weights JPEG code stream digits.

Optionally, in the storage method of the huffman table, the memory space is divided into 4 memory blocks, is respectively First memory block, the second memory block, the 3rd memory block and the 4th memory block, wherein, the first memory block memory length is 1 Code word that the code word of~4, the second memory block memory length are 5~8, the 3rd memory block memory length be 9~ The code word of 12, the code word that the 4th memory block memory length is 13~16.

Optionally, in the storage method of the huffman table, it is single that the first memory block includes up to 16 storages First, described second memory block includes up to 64 storage units, the 3rd memory block includes up to 64 storage units, described 4th memory block includes up to 64 storage units.

The present invention also provides a kind of Hofmann decoding method for JPEG, the Hofmann decoding methods for JPEG Including：

Step 10：It parses the huffman table in JPEG code streams and is subject to using the storage method of huffman table as described above Storage；

Step 20：Read in 4 JPEG code streams；

Step 30：Using 4 JPEG code streams of reading as address search huffman table, storage data are obtained；

Step 40：Judge whether the code word effective marker position stored in storage data is effective, if effectively, performing step 50；

Step 50：Code word effective length according to being stored in storage data judges the ratio for belonging to current value in current code word Special position simultaneously reads JPEG code stream digits according to the weights that storage data store, and this time Hofmann decoding terminates.

Optionally, in the Hofmann decoding method for JPEG, in step 40, if judging to obtain storage number When representing invalid according to the code word effective marker position of middle storage, then step 60 is performed；

Step 60：4 JPEG code streams are read in again, and 4 JPEG code streams of this reading are added into previous obtained storage The offset address of data storage obtains storage data, and performs step 40 as address search huffman table.

Optionally, in the Hofmann decoding method for JPEG, after step 50, further include：

Step 70：Judge JPEG code streams whether frame end, if so, decoding terminate；If it is not, step 20 is then performed again.

Optionally, in the Hofmann decoding method for JPEG,

When code word is 1~4, the operation for once reading in 4 JPEG code streams is performed；

When code word is 5~8, the operation for reading in 4 JPEG code streams twice is performed；

When code word is 9~12, the operation for reading in 4 JPEG code streams three times is performed；

When code word is 13~16, the operation of four readings, 4 JPEG code streams is performed.

In the storage method of huffman table provided by the invention and in the Hofmann decoding method of JPEG, decoding When, 4 JPEG code streams are read in every time, using 4 JPEG code streams as address search huffman table, obtain storing data, and accordingly The weights of Huffman code word are obtained, so as to complete JPEG decodings.Middle use is serial by bit decision compared with the prior art, With higher efficiency；Meanwhile only by changing the storage method of huffman table, in Hofmann decoding without using any multiple Miscellaneous logical operation, it is simply simple to read the memory of huffman table, and pass through effective marker position and go to judge code word, that is, it is not required to Complicated decoding circuit is wanted, thus also lessens the complexity of Hofmann decoding.

Description of the drawings

Fig. 1 is a kind of expression schematic diagram of storage unit of the embodiment of the present invention；

Fig. 2 is the expression schematic diagram of another storage unit of the embodiment of the present invention；

Fig. 3 is a kind of piecemeal schematic diagram of memory space of the embodiment of the present invention；

Fig. 4 is the piecemeal schematic diagram of another memory space of the embodiment of the present invention；

Fig. 5 be 5~8 bit words of the embodiment of the present invention need 64 storage units illustrate schematic diagram；

Fig. 6 is the flow diagram of the Hofmann decoding method for JPEG of the embodiment of the present invention；

Fig. 7 is the example schematic of the huffman table of a storage of the embodiment of the present invention.

Specific embodiment

Below in conjunction with the drawings and specific embodiments to the storage method of huffman table proposed by the present invention and for JPEG's Hofmann decoding method is described in further detail.According to following explanation and claims, advantages and features of the invention will It becomes apparent from.It should be noted that attached drawing is using very simplified form and uses non-accurate ratio, only to convenient, bright The purpose of the embodiment of the present invention is aided in illustrating clearly.

In Joint Photographic Experts Group, huffman coding can only be 1~16, and corresponding weights are 8, wherein high 4 weight tables Show before current value 0 number, low 4 weights represent the number of bits of current value.Based on this of huffman coding Characteristic in the embodiment of the present application, stores huffman table, the memory space is divided into 1 or multiple using a memory space Memory block, the memory space include the storage unit of multiple equal lengths, wherein, each storage unit is referred to by an address To show, the length of each storage unit is at least 11 bits, in 11 bits, 1 bit storage code word effective marker position, 2 Position bit storage code word effective length, the corresponding weights of 8 bit storage code words or offset address.

It is the structure diagram of the storage unit of the embodiment of the present application specifically, please referring to Fig.1.As shown in Figure 1, This, (each) storage unit has 11 bits, wherein, highest order represents whether the code word is effective, i.e. effective marker position；It is low 8 store the corresponding weights of the code word or offset address, which is directed toward next memory block；Intermediate 2 expressions The effective length of the code word.In the embodiment of the present application, represent that flag bit is invalid using 0,1 represents that flag bit is effective, works as storage Code word effective marker position represent it is invalid when, the code word effective length of storage be 0 namely when highest order is 0, centre two be 00.Further, when effective marker position is effective, when centre two is 00, it is 1 to represent code word size；Intermediate two are 01 When, it is 2 to represent code word size；When centre two is 10, it is 3 to represent code word size；When centre two is 11, code word size is represented For 4.

In the other embodiment of the application, each storage unit can also use other representations, for example, highest 8 Position represents to store the corresponding weights of the code word or offset address, the next memory block of offset address direction；It is 2 intermediate Represent the effective length of the code word；Lowest order represents whether the code word is effective, i.e. effective marker position, as shown in Figure 2.Further , in the other embodiment of the application, 1 expression flag bit is invalid, and 0 expression flag bit is effective etc. may be employed.

In the embodiment of the present application, the memory space is divided into 4 memory blocks, is respectively first memory block, the second storage Block, the 3rd memory block and the 4th memory block, wherein, code word that the first memory block memory length is 1~4, described the Code word that code word that two memory block memory lengths are 5~8, the 3rd memory block memory length are 9~12, the described 4th Memory block memory length is the code word of 13~16.It specifically please refers to Fig.3, is a kind of memory space of the embodiment of the present invention Piecemeal schematic diagram.In the embodiment of the present application, 4 memory blocks according to first memory block, the second memory block, the 3rd memory block and 4th memory block stores successively, can not also be specific as shown in Figure 4 in the order described above in the other embodiment of the application.

It, can be only according to the characteristic of a two field picture to be decoded when specifically carrying out Hofmann decoding to a certain two field picture There are one memory blocks, it is possibility to have two memory blocks, three memory blocks or four memory blocks.Specifically, when to be decoded one When code word in two field picture is 1~4, then only first memory block；When the code word in a two field picture to be decoded is 1~ At 8, then there are first memory block and the second memory block, and so on.

In the embodiment of the present application, the first memory block includes up to 16 storage units, second memory block most Include 64 storage units, the 3rd memory block includes up to 64 storage units, the 4th memory block includes up to more 64 storage units.The reason for each memory block needs the storage unit of above-mentioned quantity is as follows：

Huffman coding is in order to ensure decoded uniqueness, and all code words are all different prefix codes, i.e., one of code word is not It may be the prefix of another code word.If for example, have the coding that code word is 110 in code stream, then would not there are 1100 or 1101 4 bit words.So for the code word within 4, when 4 complete effective, number of codewords is most, i.e. be up to 16 codes Word, that is to say, that first memory block at most needs 16 storage units.For the code word of 5~8, can be determined due to first 4 Different prefixes, so carrying out needing 64 storage units during polishing to 1~4 below, i.e. the second memory block at most needs 64 Storage unit, can corresponding reference chart 5 to this.For the 3rd memory block and the 4th memory block, can similarly analyze, it is similary each to need Want 64 storage units.

In addition, according to Joint Photographic Experts Group, 4 kinds of huffman tables are generally comprised in JPEG code streams, are that brightness direct current, brightness are handed over respectively Stream, the exchange of colourity direct current, colourity, each huffman table are required for most 208 storage units, so a complete JPEG solution Code device at most needs 832 storage units to go to store all huffman tables.

Subsequently, the embodiment of the present application will be specifically described Hofmann decoding method for JPEG.Fig. 6 is refer to, at this Apply in embodiment, the Hofmann decoding method for JPEG includes：

First, step S10 is performed：Parse the huffman table in JPEG code streams and depositing using huffman table as described above Method for storing is stored.Wherein, existing any of which can be used for the parsing of the huffman table in JPEG code streams, The embodiment of the present application is not construed as limiting this, also repeats no more herein.

Secondly, step S20 is performed：Read in 4 JPEG code streams.

Then, step S30 is performed：Using 4 JPEG code streams of reading as address search huffman table, storage number is obtained According to.By taking huffman table shown in Fig. 3 as an example, i.e., obtain addr0, addr1 according to the 4 of reading JPEG code streams ... or An address (value) in addr207 then obtains storage data to get in into storage unit from corresponding address Hold.

Followed by execution step S40：Judge whether the code word effective marker position stored in storage data is effective.With Fig. 1 institutes Exemplified by the storage unit shown, i.e., according to the numerical value of highest order, obtain the whether effective information of code word.For example, if bit10 is effective Position represents effective, then illustrates that there are Huffman code words in current 4 JPEG code streams；Then understood by bit9 and bit8 specific Code word size.It should be noted that belonging to the bit of Huffman code word in this 4 bit address can abandon, and it is not belonging to Huffman The bit of code word must retain, because these bits belong to current value.

When obtain flag bit it is effective when, according to storage data store weights read JPEG code stream digits, that is, perform step S50.Specifically, the least-significant byte of storage data is corresponding weights, according to low 4 of weights, read from JPEG code streams corresponding Number of bits (pays attention to whether there is the bit for belonging to current value in 4 bit address, if so, being so then not belonging to from these The bit of code word starts to read), so far, current Hofmann decoding terminates, and may then continue to repeat this process.

In the embodiment of the present application, when the code word effective marker position for judging to obtain storing in storage data represents invalid, 4 JPEG code streams are read in again, the offset that 4 JPEG code streams of this reading are stored plus previous obtained storage data Location obtains storage data, that is, performs step S60 as address search huffman table；And then, step 40 is performed again：Judge Whether the code word effective marker position stored in storage data is effective.For example, if bit10 significance bits represent invalid, then represent Code word size is more than 4, needs to read in 4 JPEG code streams again at this time, and plus the preceding least-significant byte conduct once stored in data Address, goes to search stored huffman table, i.e., obtains storing data accordingly in the second memory block.Similarly, analyze again Bit10 significance bits if invalid, continue to read in 4 JPEG code streams and search the 3rd memory block and the 4th memory block, until working as Preceding decoding terminates, i.e., at most reads in 44 JPEG code streams.

In the embodiment of the present application, when code word is 1~4, the operation for once reading in 4 JPEG code streams is performed；And work as , it is necessary to perform the operation of 4 JPEG code streams of reading twice when code word is 5~8；When code word be 9~12 when, it is necessary to perform three The secondary operation for reading in 4 JPEG code streams；When code word be 13~16 when, it is necessary to perform four times reading 4 JPEG code streams operation.

From the foregoing, it will be observed that the Hofmann decoding method that the application proposes is without using the logical operation of any complexity, it is simple Reading huffman table memory, and pass through effective marker position and go to judge code word, so as to reach decoded purpose.That is the application In Hofmann decoding method there is higher efficiency, also lessen the complexity of Hofmann decoding.

It, subsequently will be with some specific JPEG code streams for the deeper Hofmann decoding method for understanding the application and proposing Exemplified by, coding/decoding method is expanded on further.

Assuming that the huffman table parsed from JPEG code streams is as shown in table 1 below：

Code word size	Code word (binary system)	Weights (hexadecimal)
			2	0 0	0x00
2	0 1	0x01
			2	1 0	0x02
3	1 1 0	0x07
			4	1 1 1 0	0x1e
5	1 1 1 1 0	0x2e

Table 1

The storage method of the huffman table proposed according to the application stores huffman table to a memory space, the storage Space is divided into first memory block and the second memory block, and specific storage form is as shown in Figure 7.

Assuming that need decoded JPEG code streams following (binary system)：

1 1 0 1 0 0 1 1 0 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1…

It is as follows to the decoding process of above-mentioned JPEG code streams：

First, 4 code streams 1101 are read in, i.e., hexadecimal 0xd goes to search the huffman table in Fig. 7 as address, obtain To storage data：(0x1,0x2,0x07) is represented in 4 data there are effective code word, and (code word is code word size for 3 110), weights 0x07 so after abandoning code word 110, reads in 7 bit binary datas of current value since the 4th： 1001101, current decoding terminates.

Then, continue to read in 4 code streams 1111, i.e., hexadecimal 0xf goes to search the Huffman in Fig. 7 as address Table obtains storage data：(0x0,0x0,0x10) represents that complete effective code word is not present in 4 code streams, i.e. code word to be more than 4, jump address 0x10, continue to read in 4 code streams 0001, i.e., hexadecimal 0x1, and add the transfer of previous read operation Address 0x10, i.e. 0x11 go to search the huffman table in Fig. 7 as address, obtain storage data：(0x1,0x0,0x2e) is represented Current 4 code streams have matched effective code word, number of significant digit 1, in addition previous 4 code streams, so complete code word is 11110, corresponding weights are 0x2e, represent there are 2 zero before current value, and the binary code of current value has 14, so The binary data of 14 is read from code stream afterwards：00111111111111, current decoding terminates.

If subsequently there is JPEG code streams, continue to decode in the manner described above；It is entire to solve if there is no JPEG code streams subsequently Code terminates.

From the foregoing, it will be observed that the advantages of Hofmann decoding method of the application proposition, is：It realizes simply, without using any complexity Logical resource；Data length per treatment fixes (i.e. 4), is easy to hardware realization；Simply carry out simple memory reading Write operation, so entire decoder can be operated in very high clock frequency.

Foregoing description is only the description to present pre-ferred embodiments, not to any restriction of the scope of the invention, this hair Any change, the modification that the those of ordinary skill in bright field does according to the disclosure above content, belong to the protection of claims Scope.

Claims (6)

1. a kind of storage method of huffman table, which is characterized in that including：Huffman table is stored using a memory space, it is described Memory space is divided into 1 or multiple memory blocks, and the memory space includes the storage unit of multiple equal lengths, wherein, often A storage unit is indicated that the length of each storage unit is at least 11 bits by an address, in 11 bits, 1 ratio Spy's storage code word effective marker position, 2 bit storage code word effective lengths, the corresponding weights of 8 bit storage code words or inclined Move address；

Wherein, in a storage unit, in 11 bits, when the code word effective marker position of storage represents invalid, storage Code word effective length for 0, and offset address for being directed to next memory block of remaining 8 storages；In a storage unit In, in 11 bits, when the code word effective marker position of storage represents effective, the code word effective length of storage, which can interpolate that, to be worked as Belong to the bit of current value in preceding code word, the corresponding weights of 8 bit storage code words simultaneously can read JPEG according to weights Code stream digit.

2. the storage method of huffman table as described in claim 1, which is characterized in that the memory space is divided into 4 storages Block, respectively first memory block, the second memory block, the 3rd memory block and the 4th memory block, wherein, the first memory block is deposited Store up length be 1~4 code word, the second memory block memory length be 5~8 code word, the 3rd memory block storage The code word that code word that length is 9~12, the 4th memory block memory length are 13~16.

3. the storage method of huffman table as claimed in claim 2, which is characterized in that the first memory block includes up to 16 A storage unit, second memory block include up to 64 storage units, the 3rd memory block includes up to 64 storages Unit, the 4th memory block include up to 64 storage units.

4. a kind of Hofmann decoding method for JPEG, which is characterized in that including：

Step 10：It parses the huffman table in JPEG code streams and uses huffman table according to any one of claims 1 to 3 Storage method stored；

Step 20：Read in 4 JPEG code streams；

Step 30：Using 4 JPEG code streams of reading as address search huffman table, storage data are obtained；

Step 40：Judge whether the code word effective marker position stored in storage data is effective, if effectively, performing step 50；

Step 50：Code word effective length according to being stored in storage data judges the bit for belonging to current value in current code word And JPEG code stream digits are read according to the weights that storage data store, this time Hofmann decoding terminates.

5. the Hofmann decoding method of JPEG to be used for as claimed in claim 4, which is characterized in that in step 40, if judging When the code word effective marker position for obtaining storing in storage data represents invalid, then step 60 is performed；

Step 60：4 JPEG code streams are read in again, and 4 JPEG code streams of this reading are added into previous obtained storage data The offset address of storage obtains storage data, and performs step 40 as address search huffman table.

6. the Hofmann decoding method of JPEG to be used for as claimed in claim 4, which is characterized in that after step 50, also wrap It includes：

Step 70：Judge JPEG code streams whether frame end, if so, decoding terminate；If it is not, step 20 is then performed again.