CN100469144C - Table transformation method for entropy encoder - Google Patents
Table transformation method for entropy encoder Download PDFInfo
- Publication number
- CN100469144C CN100469144C CN 200510029359 CN200510029359A CN100469144C CN 100469144 C CN100469144 C CN 100469144C CN 200510029359 CN200510029359 CN 200510029359 CN 200510029359 A CN200510029359 A CN 200510029359A CN 100469144 C CN100469144 C CN 100469144C
- Authority
- CN
- China
- Prior art keywords
- node
- weighted value
- incoming symbol
- brother
- symbol
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Abstract
This invention relates to entropy coder form conversion method, which comprises the following steps: processing effective change on Hageman tree according to the weighted values and its adjacent other point weighted values of coder form to improve compression efficiency; using /\ shape tree rotation and semi-rotation for effective change on Hageman tree to make input characters with current input video data to contract form time.
Description
Technical field
The invention relates to the invention of table transformation method for entropy encoder.Say in more detail, when the specific video data are compressed in to MEPG and JPEG exactly, the form of entropy coding is carried out the invention of the method for efficient transformation according to the weighted value of the weighted value of the node of incoming symbol and the node adjacent with this node.
Background technology
Along with rapid development of network, the network media form, developed into various ways such as image, video and hypertext are arranged now and deposited based on text by originally.Adopt the information of which kind of method with these forms, particularly video request program video datas such as (VOD:Video on Demand), by original analog video signal formal transformation is digital form, and storing effectively, transmit by network has then become present important problem.
Huge video data can bring serious load to network during by band-limited Network Transmission.For addressing this problem, many video data compression and backout criterion have been proposed, as: MPEG-1, MPEG-2, the MPEG-4 of ISO/IEC (InternationalOrganization for Standardzation/International Electrotechmical Commission) and ITU-T (International TelecommunicationsUnion-Telecommunication Standardization Sector) H.261 and H.263 wait.
Above-mentioned standard organically combines the lossy compression method method of quantum device and the lossless compression method of entropy coder.Particularly entropy coder utilizes the Huffman algorithm to coding video data, and this method has made full use of the probability spy of video data, can improve the compression ratio of data greatly.
Static Huffman algorithm is arranged in the Huffman algorithm.The a large amount of video data of static Huffman algorithm utilization is obtained the probability that each symbol occurs, and is the static Huffman tree of fundamental construction with the probability, again with Huffman tree as coding form to coding video data.That is, use few bit to encode, use many bits to encode the low symbol of weighted value to the high symbol of weighted value.
As shown in Figure 1a, weighted value is that external node 100 and the weighted value of 1 symbol A are that the external node 102 of 1 symbol B combines that to generate weighted value be 2 internal node 104; The internal node 104 of above-mentioned generation and weighted value are that the external node 106 of 1 symbol C combines, and generate weighted value and be 3 internal node 108; The internal node 108 of above-mentioned generation and weighted value are that the external node 110 of 2 symbol D combines, and generate weighted value and be 5 internal node 112; The internal node 112 of above-mentioned generation and weighted value are that the external node 114 of 5 symbol E combines, and generate weighted value and be 10 internal node 116, utilize these nodes to make up static Huffman tree then.
According to above-mentioned static Huffman tree, shown in Fig. 1 b, the code of formation entropy coding.That is, be that 1 symbol A is assigned as 1000 with weighted value, be that 1 symbol B is assigned as 1001 with weighted value, with weighted value is that 1 symbol A is assigned as 101, with weighted value is that 2 symbol A is assigned as 11, is that 5 symbol A is assigned as 0 with weighted value, so as to fixed video data compress and restore.
Difference according to the symbol weighted value can be built into various Huffman trees, for example shown in Fig. 2 a, according to even number symbols A, B, C, the weighted value of the external node 200,202,206,208,214 of D and E generates internal node 204,210,212,218, makes Huffman tree, again according to static Huffman tree formation entropy code shown in Fig. 2 b, again with the entropy coding that generates to fixed video data compress and restore.
Although above-mentioned static Huffman algorithm can utilize great deal of experiment data to make up the entropy coding form, the coding form that constructs can not satisfy the characteristic of various video datas, reaches the variation of conditions such as bit rate.When the video data to new input compressed, if utilize above-mentioned entropy coder, meeting was different with the symbol probability distribution of making above-mentioned Huffman tree because of the symbol probability distribution of the video data of new input, and can not effectively compress and restore.
Develop dynamic Huffman tree for the problem that solves static Huffman tree, utilize the probability distribution of present input data to rebuild an entropy coder form, the form by new structure can improve data compression rate.
Fig. 3 a, b, c are conventional dynamic Huffman tree key diagrams.For example shown in Fig. 3 a, according to even number symbols A, B, C, D, the weighted value of the external node 300,302,304,306,308,310 of E and F generates internal node 312,314,316,318,320, for example incoming symbol D, variation has taken place in probability distribution, and the dynamic Huffman tree of Fig. 3 a just can not satisfy the condition of Huffman algorithm in this case.
In this case, have symbol D external node 306 internal node 312 superior node 314 with and the weighted value of superior node 316 of external node of above-mentioned internal node 312 with symbol E of equal weight value compare; Shown in Fig. 3 b, the external node 308 of the symbol E of the internal node 316 that weighted value is high exchanges mutually with the internal node of the external node of symbol D.
Then, the superior node 318 of above-mentioned node 316 and compare with the weighted value of superior node 320 of node that above-mentioned node 316 has a symbol A of equal weight value; Shown in Fig. 3 c, external node 300 and the above-mentioned node 316 of the symbol A of the superior node 320 that weighted value is high exchange mutually.
Said process lasts till that always root node finishes again, afterwards the node 306 of symbol D and the weighted value of the node 312,316,320 under the symbol D node 306 is carried out addition respectively.
Like this, utilize dynamic Huffman tree can carry out the compression of high compression rate to new video data.When but each symbol occurs, all to the weighted value of each node be compared, make dynamic Huffman tree again, cause amount of calculation too huge, in the real-time system of reality, be difficult to use.
Summary of the invention
For addressing the above problem, the purpose of this invention is to provide a kind of table transformation method for entropy encoder, can carry out effective conversion with the weighted value of other node that is adjacent to Huffman tree according to the weighted value of the symbol node of importing the entropy coder form, thereby improve compression efficiency; Can also utilize rotation and half spinning solution of splayed tree that Huffman tree is carried out effective conversion; The weighted value of incoming symbol is had and the only value of the video data of current input; Can also shorten the table transformation time.
The object of the present invention is achieved like this: a kind of table transformation method for entropy encoder comprises:
The 10th process: the symbol of video data is input to from quantizer in the table transformation device, whether the brother of node to the node of this symbol is that external node is judged, the weighted value of the brother of node of the weighted value of the symbol node of above-mentioned input and above-mentioned input node is compared.
The 20th process: if the symbol node of input is an external node, the weighted value of input node then increases the weighted value of interdependent node less than the weighting of the brother of node of input node according to above-mentioned incoming symbol.
The 30th process: if the weighted value of the node of incoming symbol surpasses the weighted value of the brother of node of incoming symbol node, then whether the downstream site to the brother of node of the superior node of the brother of node of the superior node of incoming symbol node and incoming symbol node is that external node is judged, and the weighted value of the downstream site of the brother of node of the superior node of the weighted value of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol and incoming symbol node compared, carry out the rotation of splayed tree according to the result who judges and compare, and rebuild Huffman tree, increase the weighted value of interdependent node afterwards according to the symbol of input.
The 40th process, if the brother of node of incoming symbol node is not an external node, and the weighted value of incoming symbol node is littler than the weighted value of the brother of node of incoming symbol node, then whether the downstream site to the brother of node of the superior node of the brother of node of the superior node of incoming symbol node and incoming symbol node is that external node is judged, and the weighted value of the downstream site of the brother of node of the superior node of the weighted value of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node and incoming symbol node compared, determine to use half rotation of splayed tree according to result relatively, and rebuild Huffman tree, increase the weighted value of interdependent node then according to the symbol of input;
The 50th process is judged the processing of whether having finished all symbols of video object plane, if do not finish, then repeats since the stage 10; If finish, the value of the overall quantization parameter that generates in the use bit rate adjustment device is calculated the regulated value of weighted value; Wherein, weighted value regulated value
Wherein f () is that scope is 0.1-1.0 cellular logic reflection processor (CLIP) function, QPc is the overall quantization parameter value that is applicable to current video object face (VOP), QPp is the overall quantization parameter value of the use of front video object plane (VOP), and α and β are the constant values that tries to obtain for the difference value reflection degree of the overall quantization parameter that determines current VOP and front VOP;
The 60th process, with the regulated value and the 20th process of the weighted value that calculates, the 30th process, each node of the Huffman tree that makes up in the 40th process multiplies each other and weighted value is regulated.
Effect of the present invention:
As above explanation, table transformation method for entropy encoder of the present invention is applicable to MEPG, JPEG etc. use all video compression apparatus of Huffman algorithm, can show the current video symbolic feature well, thereby raising compression efficiency, compare form revises the required time and significantly reduces with dynamic Huffman algorithm, thereby also be applicable to real-time compressing apparatus, can keep the weighted value of each node of Huffman tree, and the symbol that obviously is dominant of output probability, thereby can distribute shorter code, improve compression efficiency.
For further specifying above-mentioned purpose of the present invention, design feature and effect, the present invention is described in detail below with reference to accompanying drawing.
Description of drawings
Fig. 1 a and Fig. 1 b are traditional static Huffman tree and code illustrated example schematic diagram.
Fig. 2 a and Fig. 2 b are other illustrated example schematic diagrames of traditional static Huffman tree and code.
Fig. 3 a is a conventional dynamic Huffman tree key diagram to Fig. 3 c.
Fig. 4 is the video compressor structural representation according to transform method of the present invention.
Fig. 5 a and Fig. 5 b are the key diagrams of common splayed tree rotation.
Fig. 6 a and Fig. 6 b are the key diagrams of common splayed tree half rotation.
Fig. 7 a and Fig. 7 b are the signal flow schematic diagrames according to form conversion method of the present invention.
Fig. 8 is a table transformation method key diagram of the present invention.
The symbol description of major part in the accompanying drawing:
400: cosine discrete transform (DCT:Discrete Consine Transform) portion forward
410: quantizer 420: memory
430: entropy coder 440: the bit rate adjustment device
450: table transformation portion 460: decipher
470: subtract and calculate device 480: diverter switch
Embodiment
4 embodiment to 8 pairs of table transformation method for entropy encoder of the present invention of accompanying drawing are elaborated below with reference to accompanying drawings.
Fig. 4 is according to video compressor structural representation of the present invention.As shown in the figure, video compressor comprises video data is carried out the forward forward cosine discrete transform portion 400 of cosine discrete transform, the quantizer 410 that the signal of the output of cosine discrete transform portion is forward quantized, for the output signal of above-mentioned quantizer being carried out entropy coding form is carried out the memory 420 that conversion stores, carry out the encoder 430 of entropy coding according to above table, bit rate according to above-mentioned entropy coder output signal is obtained the value of overall quantization parameter and the bit rate adjustment device 440 that the quantification of above-mentioned quantizer is regulated, the signal of signal of exporting in the use quantizer and the output of bit rate adjustment device carries out conversion by the inventive method to the form in the memory table transformation portion 450, the deciphering of output signal in the quantizer is the decryption part 460 of original signal, the video data of deciphering is subtracted subtracting of calculation calculate device 470, according to be interior frame (intra frame) or between frame (inter frame) video data and the above-mentioned signal that subtracts output in the calculation machine 470 that the outside receives are controlled, these signals are imported selectively the diverter switch 480 of cosine discrete transform portion 400 forward etc.
When precoding was interior frame, the variable terminal of diverter switch touched on the side fixed terminal (a).
The video data of outside input is through after the diverter switch 480, carry out forward cosine discrete transform in cosine discrete transform portion 400 forward, in quantizer 410, quantize, in entropy coder 420, above-mentioned quantitation video data are encoded again, at last output according to the form in the memory 420.
The output signal of above-mentioned entropy coder 430 is input in the bit rate adjustment device 440, the bit rate of 440 pairs of current video objects of bit rate adjustment device face is judged, generate the value of overall quantization parameter according to judged result, quantizer 410 quantizes according to the value of overall quantization parameter again, makes the signal of output from entropy coder 430 have certain bit rate.
Signal input table transformation portion 450 with output in above-mentioned quantizer 410 and the bit rate adjustment device 430, table transformation portion 450 becomes the memory form according to table transformation method of the present invention with above-mentioned conversion of signals, and entropy coder 430 carries out entropy coding according to the form that is transformed into again.
Video data input decipher 460 after will quantizing then is transformed into original video data, is input to subtract again and calculates in the device 470, detects the difference of the video data of these data and outside input.
When coding was frame, the convertible tip of diverter switch 480 touched on other the side fixed terminal (b).
Being input to forward by diverter switch 480 the above-mentioned difference that subtracts calculation machine 470 detected video datas, cosine discrete transform portion 400 carries out forward cosine discrete transform, be input to then in the quantizer 410 and quantize, the entropy coder form that is transformed into according to foundation transform method of the present invention in entropy coder carries out entropy coding again, then output.
Coding be under the situation of a frame in situation during frame identical, bit rate adjustment device 440 generates the value of overall quantization parameter according to the output signal of entropy coder 430, quantizer 410 quantizes according to the value of overall quantization parameter again, and the signal according to output in above-mentioned quantizer 410 and the bit rate adjustment device 430 carries out conversion to the coding form in the memory according to method of the present invention in entropy coder then.
Transform method of the present invention uses rotation of splayed tree and half rotation.Fig. 5 is common splayed tree rotation key diagram.Shown in Fig. 5 a, at symbol A, B, the external node (500 of C node, 502,504) and the external node (500 of symbol A and B, 502) incoming symbol A in the Huffman tree of internal node of Xing Chenging (506) and the internal node 508 that forms by the external node 504 of above-mentioned internal node 506 and symbol C, under the situation that the weighted value of external node 500 increases, shown in Fig. 5 b, the internal node 506 of external node 500 is transformed into the superior node of internal node 508, and with the brother of node of the external node 500 of symbol A, that is, weighted value does not have the external node 502 of the symbol B of variation to be transformed into the brother of node 500 of symbol C; If the weighting of symbol C increases among Fig. 5 b, then change as Fig. 5 a.
That is, input institute decide symbol, when the weighted value of node changes, by the rotation of splayed tree, when the mutual alignment of internal node takes place to change, the brother of node of importing external node is placed on the position of other internal node.
Fig. 6 is that common splayed tree is partly rotated key diagram.Shown in Fig. 6 a, incoming symbol A, when the weighted value of external node 600 changes, then shown in Fig. 6 b, the invariant position of internal node 606,608, the position of conversion external node 600 and external node 604: incoming symbol C shown in Fig. 6 b, when the weighted value of external node 604 changes, then shown in Fig. 6 a the external node 604,600 of symbol A and C mutual switch.
How to utilize as shown in Figure 5 and Figure 6 rotation of splayed tree and half rotation to describe below with reference to Fig. 7 and Fig. 8 to the present invention.
Shown in Fig. 7 a, in the stage 700 with fixed symbol from quantizer, be input to table transformation portion, in the stage 702, determine the brother of node of above-mentioned symbol node then.As Fig. 8, during incoming symbol X, the brother of node 802 of the external node 800 of symbol X is confirmed.
After in stage 702 brother of node 802 being confirmed, in the next stage 704, be that external node or internal node are judged to the above-mentioned brother of node 802.Be the above-mentioned brother of node 802 whether be from quantizer 410 input decide symbol node judge.
Judged result according to the stage 704 compared the symbol node 800 of above-mentioned input and the weighted value of its brother of node 802 in stage 706 and stage 708 then.
In the stage 704, judge above-mentioned node 802 be from quantizer 410 inputs fixed symbol node promptly be external node, and when relatively result is Wx＜Wy in the stage 706, do not change the position of each node, and in the stage 738, be that the weighted value of external node 800 and internal node 804,806 is regulated to corresponding node according to the symbol X of above-mentioned input.
When the stage 706, when the comparative result in 708 stages is Wx 〉=Wy, no matter the result who judges in the stage 704 how, then the brother of node 808 to the superior node 804 of above-mentioned node 800 is confirmed in the stage 710, in the stage 712, whether be that external node is judged to node 808, according to the result who judges in the stage 714,716 to the weighted value Wx of said external node 800 and node 808, Wz compares.
Judged result when the stage 712 is that node 808 is external nodes, and the comparative result in stage 714 is when being Wx＜Wz, do not change the position of each node, symbol X according to above-mentioned input regulates the weighted value of corresponding external node 800 and internal node 804,806 in the stage 738 then.When the stage 714, when 716 comparative result is Wx 〉=Wz, no matter the judged result in stage 712 how, is used the rotation of splayed tree in the stage 718, symbol X according to above-mentioned input regulates with the weighted value of corresponding external node 800 and internal node 804,806 in the stage 738 then.
If the judged result in stage 712 is a node 808 is internal nodes, and the comparative result in stage 716 is when being Wx＜Wz, downstream site 810 to above-mentioned node 808,812 weighted value (Wz1, Wz2) confirm, in the stage 722, whether Wx is judged greater than Wz1 in the stage 724,726, whether Wx is judged greater than Wz2.
If the result who judges is Wx〉Wz1, Wx〉Wz2, then in the stage 728, use the rotation of splayed tree, the symbol X according to above-mentioned input regulates the weighted value of corresponding external node 800 and internal node 804,806 in the stage 738 then.
If the judged result in stage 722 is Wx〉Wz1, and the judged result in stage 724 is not Wx〉Wz2; Or the judged result in stage 722 is not Wx〉Wz1 and the judged result in stage 726 is Wx Wz2, then in the stage 730 to above-mentioned node 810,812 which weighted value are judged greatly, in the stage 732, whether be that external node is judged then to the big node of weighted value, then in the stage 728, use the rotation of splayed tree if not external node, symbol X according to above-mentioned input regulates the weighted value of corresponding external node 800 and internal node 804,806 in the stage 738 then.If the judged result in stage 732 is external node then the position of not changing each node, the symbol X according to above-mentioned input in the stage 738 regulates the weighted value of corresponding external node 800 and internal node 804,806.
If the judged result in stage 722 is not wX〉wZ1, the judged result in stage 726 is not wX〉wZ2, then in the stage 734 to above-mentioned node 810, whether exist external node to judge in 812, if have external node then use splayed tree rotation in the stage 736, symbol X according to above-mentioned input regulates the weighted value of corresponding external node 800 and internal node 804,806 in the stage 738 then; If there is no external node does not then change the position of each node, and the symbol X according to above-mentioned input in the stage 738 regulates the weighted value of corresponding external node 800 and internal node 804,806.
If the judged result in stage 704 is that above-mentioned node 802 is internal nodes, when the comparative result in stage 708 is Wx＜Wy, shown in Fig. 7 b, the brother of node 808 to the superior node 804 of above-mentioned node 800 in the stage 740 is confirmed, in the stage 742, whether be that external node is judged to node 808, according to the result who judges in the stage 744,746 to node 800,808 weighted value Wx, Wy compares.
Judged result when the stage 742 is that node 808 is not an external node, when the comparative result in stage 744 is Wx＜Wz, do not change the position of each node, the symbol X according to above-mentioned input in the stage 768 regulates the weighted value of corresponding external node 800 and internal node 804,806.If the comparative result in stage 744,746 is Wx 〉=Wz, in the stage 748, use half rotation of splayed tree, the symbol X according to above-mentioned input regulates the weighted value of corresponding external node 800 and internal node 804,806 in the stage 768 then.
Judged result when the stage 742 is that node 808 is external nodes, when the comparative result in stage 746 is Wx＜Wz, in the stage 750, downstream site 810 to above-mentioned node 808, (Wz1 Wz2) confirms 812 weighted value, judges it is whether Wx is greater than Wz1 in the stage 752, judge that in the stage 754,756 whether Wx is greater than Wz2.
When judged result is Wx〉Wz1, Wx〉during Wz2, in the stage 758, use half rotation of splayed tree, the symbol X according to above-mentioned input in the stage 768 regulates the weighted value of corresponding external node 800 and internal node 804,806.
Judged result when the stage 752 is Wx〉Wz1, the judged result in stage 754 is not Wx〉Wz2 or the judged result in stage 752 be not Wx〉Wz1, the judged result in stage 754 is Wx〉during Wz2, in the stage 760, judge above-mentioned node 810, the weighted value of which node is bigger in 812, judges in the stage 762 whether the big node of weighted value is external node.If not external node, then in the stage 758, use half rotation of splayed tree, the symbol X according to above-mentioned input regulates the weighted value of corresponding external node 800 and internal node 804,806 in the stage 768 then.If the judged result in stage 762 is to be external node then the position of not changing each node, the symbol X according to above-mentioned input in the stage 768 regulates the weighted value of corresponding external node 800 and internal node 804,806.
If the judged result in stage 752 is not Wx〉Wz1, the judged result in stage 746 is not Wx〉Wz2, then in the stage 764 to above-mentioned node 810, whether exist external node to judge in 812, if have external node then use splayed tree half rotation in the stage 766, symbol X according to above-mentioned input regulates the weighted value of corresponding external node 800 and internal node 804,806 in the stage 768 then; If judged result does not then change the position of each node for there not being external node, the symbol X according to above-mentioned input in the stage 768 regulates the weighted value of corresponding external node 800 and internal node 804,806,
Through the above-mentioned stage, after Huffman tree is made up again and having increased weighted value according to the node of input, in the stage 770, handle and judge, if the result who judges is for all not finishing then turn back to the stage 700 and repeat whether all finishing the symbol that constitutes video object plane (VOP).
If the judged result in stage 770 regulated value to the weighted value of each node that will regulate Huffman tree for all finishing then in the stage 772 calculates, in the stage 774, the weighted value of each node of constituting Huffman tree is regulated according to the weighted value regulated value that calculates.
Promptly, image height definition television (HDTV:high definition television) is imported under the situation of unlimited continuous video data like that, the weighted value of accumulating on each node of Huffman tree does not only have the memory space that can enough store, and can not reflect the situation of change of weighted value.For example: when the probability of special symbol obviously is better than other symbol, give the coding of its 1BIT according to the Huffman algorithm assigns, on the contrary, if it is too many to have the symbol of equal probabilities, because mutually combining between the symbol, bit number to its distribution will become many, only utilizes the weighted value structure Huffman tree of the video data of input at present in this case, because the minimizing of the chance that the symbol that probability obviously is dominant occurs can reduce the efficient of coding on the contrary.Therefore, should carry out suitable adjusting with holding time to the size of weighted value, the chance that the symbol that the raising probability obviously is dominant occurs is so that present video changes direct structure generation effect to Huffman tree.
Therefore, according to the present invention, the value of the overall quantization parameter of each video object plane of decision (VOP) in the bit rate adjustment device, in the stage 772, use 1 pair of weighted value regulated value of formula to calculate then, in the stage 774, multiply each other again, again the weighted value of each node is regulated with the weighted value of each node of Huffman tree.
[formula 1]
The weighted value regulated value
Wherein f () is that scope is 0.1-1.0 cellular logic reflection processor (CLIP) function, and QPc is the overall quantization parameter value that is applicable to current video object face (VOP), and QPp is the overall quantization parameter value that front video object plane (VOP) uses.
α and β are the constant values that tries to obtain for the difference value reflection degree of the overall quantization parameter that determines current VOP and front VOP, and in the transform method of the present invention, by a large amount of tests, the α value generally is fixed on 0.5, and the β value generally is fixed on 0.3.
In the above-mentioned formula 1, the difference of the QP value of current VOP and front VOP is big more, and the conversion of picture is just rapid more, and it is just high that brightness changes the probability that takes place, just big more with the difference of previous symbol probability.Because the weighted value of front reduces greatly, encodes with regard to the weighted value of using new incoming symbol.If both QP value differences are different little, then the front weighted value is still kept initial value, for the difference that increases between each symbol offers an opportunity.
Those of ordinary skill in the art will be appreciated that, above embodiment is used for illustrating the present invention, and be not to be used as limitation of the invention, as long as in connotation scope of the present invention, all will drop in the scope of claims of the present invention variation, the modification of the above embodiment.
Claims (5)
1, a kind of table transformation method for entropy encoder is characterized in that comprising following process:
The 10th process: the symbol of video data is input to from quantizer in the table transformation device, whether the brother of node to the node of this symbol is that external node is judged, the weighted value of the brother of node of the weighted value of the symbol node of above-mentioned input and above-mentioned input node is compared;
The 20th process: if the symbol node of input is an external node, the weighted value of input node then increases the weighted value of interdependent node less than the weighting of the brother of node of input node according to above-mentioned incoming symbol;
The 30th process: if the weighted value of the node of incoming symbol surpasses the weighted value of the brother of node of incoming symbol node, then whether the downstream site to the brother of node of the superior node of the brother of node of the superior node of incoming symbol node and incoming symbol node is that external node is judged, and the weighted value of the downstream site of the brother of node of the superior node of the weighted value of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol and incoming symbol node compared, carry out the rotation of splayed tree according to the result who judges and compare, and rebuild Huffman tree, increase the weighted value of interdependent node afterwards according to the symbol of input;
The 40th process, if the brother of node of incoming symbol node is not an external node, and the weighted value of incoming symbol node is littler than the weighted value of the brother of node of incoming symbol node, then whether the downstream site to the brother of node of the superior node of the brother of node of the superior node of incoming symbol node and incoming symbol node is that external node is judged, and the weighted value of the downstream site of the brother of node of the superior node of the weighted value of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node and incoming symbol node compared, determine to use half rotation of splayed tree according to result relatively, and rebuild Huffman tree, increase the weighted value of interdependent node then according to the symbol of input;
The 50th process is judged the processing of whether having finished all symbols of video object plane, if do not finish, then repeats since the stage 10; If finish, the value of the overall quantization parameter that generates in the use bit rate adjustment device is calculated the regulated value of weighted value; Wherein, weighted value regulated value
Wherein f () is that scope is 0.1-1.0 cellular logic reflection processor (CLIP) function, QPc is the overall quantization parameter value that is applicable to current video object face (VOP), QPp is the overall quantization parameter value of the use of front video object plane (VOP), and α and β are the constant values that tries to obtain for the difference value reflection degree of the overall quantization parameter that determines current VOP and front VOP;
The 60th process, with the regulated value and the 20th process of the weighted value that calculates, the 30th process, each node of the Huffman tree that makes up in the 40th process multiplies each other and weighted value is regulated.
2, table transformation method for entropy encoder as claimed in claim 1 is characterized in that described the 30th process comprises following process:
Whether the 31st process is that external node is judged to the brother of node of the superior node of incoming symbol node, and the weighted value of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node is compared;
The 32nd process, if the brother of node of the superior node of incoming symbol node is an external node, the weighted value of incoming symbol node increases the weighted value of interdependent node than the weighted value of the brother of node of the superior node of incoming symbol node hour according to incoming symbol;
The 33rd process, if when the weighted value of incoming symbol node surpasses the weighted value of the brother of node of superior node of incoming symbol node, the rotation of splayed tree is used in decision, rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 34th process, if the brother of node of the superior node of incoming symbol node is not an external node, the weighted value of incoming symbol node is during less than the weighted value of the brother of node of the superior node of incoming symbol node, whether the downstream site to the brother of node of the superior node of incoming symbol node is that external node is judged, and the weighted value of the downstream site of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node compared, whether decision uses the rotation of splayed tree according to comparative result, and rebuild Huffman tree, increase the weighted value of interdependent node then according to incoming symbol.
3, table transformation method for entropy encoder as claimed in claim 2 is characterized in that described the 34th process comprises following process:
The 34-1 process compares the weighted value of two downstream sites of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node;
The 34-2 process, if the weighted value of incoming symbol node is all bigger than the weighted value of two downstream sites, then the rotation of splayed tree is used in decision, and rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 34-3 process, littler if the weighted value of incoming symbol node is bigger than one of them of two downstream sites than another, whether be that external node is judged then to the big downstream site of weighted value;
The 34-4 process, if the big downstream site of weighted value is not an external node, then the rotation of splayed tree is used in decision, and rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 34-5 process if the big downstream site of weighted value is an external node, then increases the weighted value of interdependent node according to incoming symbol;
Whether the 34-6 process is if the weighted value of incoming symbol node is all littler than the weighted value of two downstream sites, then to existing external node to judge in two downstream sites;
The 34-7 process, if there is external node, then the rotation of splayed tree is used in decision, and rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 34-8 process, if there is no external node then increases the weighted value of interdependent node according to incoming symbol.
4, table transformation method for entropy encoder as claimed in claim 1 is characterized in that described the 40th process comprises following process:
Whether the 41st process is that external node is judged to the brother of node of the superior node of incoming symbol node, and the weighted value of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node is compared;
The 42nd process, if the brother of node of the superior node of incoming symbol node is an external node, and the weighted value of incoming symbol node then increases the weighted value of interdependent node less than the weighted value of the brother of node of the superior node of incoming symbol node according to incoming symbol;
The 43rd process, if the weighted value of incoming symbol node is above the weighted value of the brother of node of the superior node of incoming symbol node, then half rotation of splayed tree is used in decision, and rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 44th process, if the brother of node of the superior node of incoming symbol node is not an external node, and the weighted value of incoming symbol node is less than the weighted value of the brother of node of the superior node of incoming symbol node, then whether the downstream site to the brother of node of the superior node of incoming symbol node is that external node is judged, and the weighted value of the downstream site of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node compared, whether use half rotation of splayed tree according to result's decision of judging and comparing, and rebuild Huffman tree, increase the weighted value of interdependent node then according to incoming symbol.
5, table transformation method for entropy encoder as claimed in claim 4 is characterized in that described the 44th process comprises following process:
The 44-1 process compares the weighted value of two downstream sites of the brother of node of the superior node of the weighted value of incoming symbol node and incoming symbol node;
The 44-2 process, if the weighted value of incoming symbol node is all bigger than the weighted value of two downstream sites, then half rotation of splayed tree is used in decision, and rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 44-3 process, littler if the weighted value of incoming symbol node is bigger than one of them of two downstream sites than another, whether be that external node is judged then to the big downstream site of weighted value;
The 44-4 process, if the big downstream site of weighted value is not an external node, then half rotation of splayed tree is used in decision, and rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 44-5 process if the big downstream site of weighted value is an external node, then increases the weighted value of interdependent node according to incoming symbol;
Whether the 44-6 process is if the weighted value of incoming symbol node is all littler than the weighted value of two downstream sites, then to existing external node to judge in two downstream sites;
The 44-7 process, if there is external node, then half rotation of splayed tree is used in decision, and rebuilds Huffman tree, increases the weighted value of interdependent node then according to incoming symbol;
The 44-8 process, if there is no external node then increases the weighted value of interdependent node according to incoming symbol.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200510029359 CN100469144C (en) | 2005-09-02 | 2005-09-02 | Table transformation method for entropy encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200510029359 CN100469144C (en) | 2005-09-02 | 2005-09-02 | Table transformation method for entropy encoder |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1925613A CN1925613A (en) | 2007-03-07 |
| CN100469144C true CN100469144C (en) | 2009-03-11 |
Family
ID=37818034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200510029359 Expired - Fee Related CN100469144C (en) | 2005-09-02 | 2005-09-02 | Table transformation method for entropy encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100469144C (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2734980B1 (en) | 2011-07-18 | 2018-03-21 | Thomson Licensing | Method for adaptive entropy coding of tree structures |
-
2005
- 2005-09-02 CN CN 200510029359 patent/CN100469144C/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN1925613A (en) | 2007-03-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1316433C (en) | Video-information encoding method and video-information decoding method | |
| CN101242531B (en) | Video coding system providing separate coding chains for dynamically selected small-size or full-size playback | |
| CN101420614B (en) | Image compression method and device integrating hybrid coding and wordbook coding | |
| CN100596198C (en) | Encoding method and encoder | |
| CN103404141A (en) | Video encoding and decoding using transforms | |
| CN101366282B (en) | Method and apparatus for encoding and decoding video signals on group basis | |
| US9307247B2 (en) | Compression using range coding with virtual sliding window | |
| CN1825976A (en) | Method of decoding a digital video sequence and related apparatus | |
| CN106937112A (en) | Bit rate control method based on H.264 video compression standard | |
| CN106612439A (en) | Adaptive fast fractal image compression method | |
| US10110896B2 (en) | Adaptive motion JPEG encoding method and system | |
| CN100469144C (en) | Table transformation method for entropy encoder | |
| JPH0787491A (en) | Quantization device, inverse quantization device, picture processing unit and quantization method, reversed quantization method and picture processing method | |
| CN102256126A (en) | Method for coding mixed image | |
| Arora et al. | Review of Image Compression Techniques | |
| CN101198056A (en) | Variable length encoding method and device | |
| CN102307303A (en) | Ternary-representation-based image predictive coding method | |
| CN101212543A (en) | Image quality and constrained code rate based JPEG2000 code rate control method | |
| CN101047857B (en) | Method for implementing layering variable-length code decode | |
| US6125211A (en) | Progressive image transmission | |
| US20060222245A1 (en) | Image encoding apparatus | |
| Kanefsky et al. | Predictive source coding techniques using maximum likelihood prediction for compression of digitized images | |
| KR100389702B1 (en) | The method for compressing and expanding data-stream without loss by bit-substitution | |
| Zhang et al. | Context-based arithmetic coding reexamined for DCT video compression | |
| CN105007483B (en) | A kind of screen content coding-decoding method compatible with H264 standards |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090311 Termination date: 20091009 |