CN101184227A - Deblocking filter - Google Patents
Deblocking filter Download PDFInfo
- Publication number
- CN101184227A CN101184227A CNA2006101387568A CN200610138756A CN101184227A CN 101184227 A CN101184227 A CN 101184227A CN A2006101387568 A CNA2006101387568 A CN A2006101387568A CN 200610138756 A CN200610138756 A CN 200610138756A CN 101184227 A CN101184227 A CN 101184227A
- Authority
- CN
- China
- Prior art keywords
- filtering
- huge square
- cells
- block
- limit
- 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.)
- Pending
Links
Images
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The invention relates to a block-removing filter, which comprises two sets of filter calculation circuits, a transposition buffer used for transferring data filtering direction, an output caching area for middle value and final result during storage calculation process, and a control and output module used for judging various combinations and responsible for processing external input and output data. The invention is applicable for filter framework that the square block of pure frame image or pure field image coding or both is adjustable, moreover by the configuration of internal buffer and the control of filter flow process, the filtering time and the circuit area can be lessened effectively.
Description
Technical field
The invention relates to a kind of images filter device framework, be particularly to a kind of filter framework that is applicable to H.264 main level and high-level.
Background technology
At present newer video compression technology H.264, among the MPEG4, all add deblocking filter (De-blocking filter), its role is to improve the compression ratio of encoder, and after decoding, picture is removed the smothing filtering of block, effectively promote the quality of output picture.
According to standard H.264/AVC, filter is respectively to take out 4 pixels (pixel) about the border of filtering when carrying out filtering, carries out the action of filtering.Shown in the known deblocking filter filtering action of Fig. 1 schematic diagram, the order of filtering must be finished the filtering of horizontal direction earlier, then carries out the filtering of vertical direction.If a huge square is divided into 16 block of cells,, will carry out four sub-level trend pass filterings and the action of finishing whole filtering is just calculated in four vertical direction filtering with the block of cells of 4 * 4 (pixel).
In order to finish above-mentioned action, and take into account the demand on the speed, in more existing designs, the main part of inquiring into is the order by design filtering mostly, makes vertical filtering action with horizontal direction to carry out simultaneously; Secondly because when carrying out filtering, therefore each 4 * 4 block of cells all must must set up some buffer buffering areas through the action of four filtering, store and only finish the partly block of filtering action.If therefore can finish the complete filtering action of a block more soon in design, the then required buffering area of using is few more, and the area of design is also more little, and also can reach this requirement by good filter sequence.
Seeing also Fig. 2, be known deblocking filter filter sequence schematic diagram, is more common filter sequence in the existing at present design.The square of top maximum is the huge square of a brightness (Luma), below less then be the huge square of corresponding colourity (Chroma).Wherein with one 4 * 4 block of cells as the cutting unit, each block of cells wants the border of filtering to have four, every borderline digitized representation the order of filtering.Owing to will observe and do horizontal direction earlier, just do the rule of vertical direction, by the block of cells in the upper left corner on the figure, can find out article one border of vertical direction, two borders must waiting until horizontal direction all filtering finish after could begin to carry out.It is also identical that same numbers is then represented the time of carrying out filtering on the border.
Yet the framework of known deblocking filter can only be handled the huge square (Macroblock) that only comprises hardwood figure (Frame) or field pattern (Filed) coding among the figure, is at H.264 main level is designed; But in high-level H.264, the huge square in same figure, its hardwood, a coding can be along with each huge square changes (Macroblock pair), so the framework of known deblocking filter is to be suitable for.Because each huge square when carrying out filtering, also must take out 4 pixels of boundary member from the huge square on top and the left side simultaneously.If therefore hardwood, a coding can change with huge square, then when the filtering action of carrying out intersection, under different boundary conditions, filtering can be different with pure hardwood figure or pure field pattern coding, therefore must consider the boundary condition that these are different at design aspect.
Summary of the invention
In view of this, the present invention provides one applicable to pure hardwood figure or pure field pattern, but also or hardwood, the filter framework when field pattern is encoded huge square modulation (Macroblock-Adaptive Frame-Field Coding).And the present invention sees through the design of filtering framework, and the control of filtering flow process, can reach the effect of dwindling circuit area and shortening the filtering time.
The invention provides a kind of deblocking filter in order to achieve the above object, it is characterized in that, comprise a horizontal filter, be to be used to receive a huge square, and this huge square is carried out the filtering of horizontal direction; One vertical filter is to be used for this huge square is carried out the filtering of vertical direction; One output buffer is to be connected between this horizontal filter and this vertical filter, is used for temporary through filtered this huge square of horizontal direction and through filtered this huge square of vertical direction; One transpose buffering district, be to be connected in this horizontal filter, this vertical filter and this output buffer, receive the block of cells of having finished filtered this huge square of horizontal direction, the block of cells transposition that is used for finishing filtered this huge square of horizontal direction carries out the filtering of vertical direction for this vertical filter; One Data Management Unit is to be connected in output state, in order to judge the encoding state of this huge square, reaches this huge square that horizontal filtering and vertical filtering are finished in output; An and control unit, be to be connected in this horizontal filter, this vertical filter, this transpose buffering district and this Data Management Unit, in order to control this horizontal filter, this vertical filter, this transpose buffering district and the execution of this Data Management Unit filtering action to this huge square.
The present invention proposes a kind of blocking filtering method of removing in addition, it is characterized in that, the required huge square in a top when at first reading one first huge square filtering; Then carry out the filtering of one first huge square, read the required huge square in top of one second huge square simultaneously; Carry out the filtering of one second huge square then, the result with this first huge square filtering writes back in the DRAM (Dynamic Random Access Memory) simultaneously; Filtering result with this second huge square writes back in this DRAM (Dynamic Random Access Memory) at last.
Description of drawings
Fig. 1 is known deblocking filter filtering action schematic diagram;
Fig. 2 is known deblocking filter filter sequence schematic diagram;
Fig. 3 is deblocking filter framework Data Control flow chart of the present invention;
Fig. 4 be for deblocking filter of the present invention be the system Organization Chart;
Fig. 5 is the allocation plan for output buffer of the present invention;
Fig. 6 is the allocation plan for transpose buffering of the present invention district;
Fig. 7 is that present huge square is that hardwood graph code, the huge square in top are the filter sequence schematic diagram of field pattern coding in the block ripple device for the present invention removes
Fig. 8 is the filter sequence schematic diagram under the deblocking filter normal condition of the present invention.
Symbol description
Transpose buffering district 105
Buffer 1051
First temporary storage buffer region 1061
Second temporary storage buffer region 1062
Direct memory access unit 110
DRAM (Dynamic Random Access Memory) 2
Block of cells A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, V, W, X
Embodiment
(Cb, Cr), and each huge square is when carrying out filtering, simultaneously also must be from the top and the huge square on the left side 4 pixels of taking out boundary member because one group of huge square has comprised a huge square of brightness (Y) and two huge squares of colourity.Images filter device framework of the present invention needs to read to DRAM (Dynamic Random Access Memory) the data of the huge square in top when the filtering of doing present huge square, and the required data of last huge square is kept in, and uses for present huge square filtering.
See also Fig. 3 deblocking filter framework of the present invention Data Control flow chart, the step of images filter device framework overall data control flow of the present invention at first is to be ready to back (when being meant that the required filtering data of present huge square has been stored in the buffer in the huge square in the left side) when last huge square, just start from reading in the DRAM (Dynamic Random Access Memory) brightness square required huge square pixel in top (as Fig. 3 step S301) when filtering.When filtering, behind the required huge square in top,, in DRAM (Dynamic Random Access Memory), read the huge square pixel in the required top of the huge square of colourity (as Fig. 3 step S303) simultaneously and run through the brightness square just carry out the filtering of brightness square.Then after the filtering of finishing the brightness square and reading the huge square pixel in the required top of the huge square of colourity, just begin the filtering of colourity square, the result with the filtering of brightness square writes back (as Fig. 3 step S305) in the DRAM (Dynamic Random Access Memory) simultaneously.At last after the filtering of the huge square of colourity is finished, the result of the huge square of colourity is write back (as Fig. 3 step S307) in the DRAM (Dynamic Random Access Memory).
Above-mentioned flow process control can be carried out reading of colourity square to DRAM (Dynamic Random Access Memory) simultaneously when carrying out brightness and colourity filtering, reach writing back of brightness square.Because the access of dynamic randon access memory is not timely, send requirements (Request) after, need line up with other module, after a while back just can obtain answer.Therefore the main cause of doing like this is by being hidden in the process of filtering the time of delay with the dynamic randon access memory, to shorten the required time of whole filtering.
Then see also Fig. 4, be for deblocking filter of the present invention be the system Organization Chart.Deblocking filter 1 of the present invention is to comprise control unit 101, input buffer 102, horizontal filter 103, vertical filter 104, transpose buffering district 105, output buffer 106, Data Management Unit 107, data processing unit 108, computing comparing unit 109 and direct memory access unit 110.Wherein direct memory access unit 110 is the addresses that produce DRAM (Dynamic Random Access Memory) 2, for Data Management Unit 107 data is write DRAM (Dynamic Random Access Memory) 2.Input buffer 102 is to be used for the temporary huge square that will read from DRAM (Dynamic Random Access Memory) 2, comprises huge square of brightness and the huge square of colourity.After the huge square of brightness and the huge square of colourity have been ready to, just offer the filtering that horizontal filter 103 begins to do horizontal direction, and when carrying out brightness and colourity filtering, simultaneously DRAM (Dynamic Random Access Memory) 2 is carried out reading of colourity square.And data processing unit 108 is the encoding states that are used to detect adjacent huge square, as detects the huge square that adjacent huge square is hardwood figure (Frame) or field pattern (Filed) coding, or its hardwood, a changeable huge square of coding are to (Macroblock pair) etc.; And computing comparing unit 109 is the processing that are used for huge square parameter, as the comparison of Alpha, Beta and bs value, and then can allow 101 pairs of various different situations combinations of control unit make corresponding filtering control.
And deblocking filter 1 uses filter (Filter) counting circuit of two covers in order to carry out the filtering action of level and vertical direction with the time on the framework, and horizontal filter 103 and vertical filter 104 are done filtering; Transpose buffering district 105 (Transpose Buffer) is used for translation data filtering direction; By median and the final result in the output buffer 106 storage computational processes; And utilize control unit 101, Data Management Unit 107, computing comparing unit 109 and data processing unit 108 to judge that various different situations make up and are responsible for outside input, output data are handled.
In addition for fear of 2 accesses of unnecessary DRAM (Dynamic Random Access Memory), to improve the utilization ratio of DRAM (Dynamic Random Access Memory) 2, during the judgement of control unit 101 before carrying out filtering, if the huge square that will carry out filtering at present, its the top or Far Left do not need to carry out filtering, then when filtering, can skip over and read the brightness square, not read the data of the huge square in top to DRAM (Dynamic Random Access Memory) 2 with the action of reading the colourity square; In addition after the filtering of at every turn finishing one group of huge square, just directly next being organized data that huge square will carry out filtering directly deposits and is arranged in output buffer 106, when descending the huge square filtering of group, just need not read the data of the huge square in the left side again to DRAM (Dynamic Random Access Memory) 2 like this.The average time of filtering also can be significantly shortened in top these two designs of carrying.
See also Fig. 5 and be allocation plan for output buffer of the present invention.The configuration of output buffer 106 comprises first temporary storage buffer region 1061, second temporary storage buffer region 1062 and buffer array 1063 (register array).Top buffer array 1063 is used for storing when carrying out the top boundary filtering, the huge square pixel in required top of using, and the left side second temporary storage buffer region 1062 is used for storing adjacent huge square pixel required when carrying out the Far Left boundary filtering.The adjacent huge square quantity of required storage, when H264 main stor(e)y level, but owing to there is not the encode consideration of huge square modulation (Macroblock-Adaptive Frame-Field Coding) of hardwood, field pattern, therefore only need use the temporary storage buffer region of 4 * 16 * 8 (positions), the mode that reads is also more simple.And in the H264 high-level, but if hardwood, field pattern are encoded during huge square modulation, the huge square that transmits is paired, the quantity that therefore will store just becomes original twice 8 * 16 * 8 (position), the mode of access also can be along with the coded system of present huge square and adjacent huge square different and change to some extent.And first temporary storage buffer region 1061 is the output buffers for present huge square.First temporary storage buffer region 1061 and second temporary storage buffer region 1062 are to be a static random access memory.
See also Fig. 6 and be allocation plan for transpose buffering of the present invention district.In deblocking filter 1 calculation process, can deposit block value in first temporary storage buffer region 1061 through horizontal direction filtering for the first time, then to take out again and carry out secondary horizontal filtering action, the value after finishing deposits in the transpose buffering district 105.This transpose buffering district 105 is made up of the buffer 1051 of 7 group 4 * 4 * 8 (position), and the buffer 1051 of each group 4 * 4 * 8 can be used to store a block of cells of finishing horizontal direction filtering, and reads the filtering of carrying out other direction by vertical direction.By the time the filtering of vertical direction also finish after, write back again at last in first temporary storage buffer region 1061.The configuration frame figure that Fig. 6 the right is shown as 4 * 4 * 8 buffer 1051 when the block of cells of finishing horizontal direction filtering, is just imported the block of cells of finishing, again by the output of first output, to do the filtering of vertical direction by first input end.And finish when the filtering of vertical direction, just by the input of second input, read by second output again, with correct writing back in first temporary storage buffer region 1061.Adopt such way to reduce the number of buffer to dwindle the area of circuit.And after all block of cells are all finished filtering in first temporary storage buffer region 1061, just can write in the DRAM (Dynamic Random Access Memory) 2 through the huge square data that Data Management Unit 107 will be handled well, and this action is can work as the colourity square when filtering, to write back in the DRAM (Dynamic Random Access Memory) 2 with the brightness square that filtering is finished simultaneously, as above control flow is described.
And the input and output of above-mentioned each buffer 1051 are all only from fixing buffer input and output, coiling complexity and circuit area have been reduced, but and when encoding huge square modulation at hardwood, field pattern, data read and writing mode have also been made corresponding change and design, it is very smooth and easy to make that whole framework all can operate when but general situation is encoded huge square modulation with hardwood, field pattern, can not waste the extra buffer and the time of filtering.
So the framework in output buffer 106 and transpose buffering district 105 is all suitable in H264 main stor(e)y level and high-level, only data read, writing mode are comparatively complicated when high-level, can be according to the coded system of present huge square and adjacent huge square different have different.
And the action of the data access of output buffer 106 is the hardwood graph code at present huge square, when the huge square in top is the field pattern coding, is with different generally speaking.As shown in Figure 5, in order to reduce coiling and the complexity of controlling, every group of buffer array 1063 that is used for the storage cell piece, its input and output in all directions all are from same group of buffer, therefore can not specify wherein wantonly four buffers to go access data wherein arbitrarily.Under general situation, the data acess method of removing block ripple device 1 is the coded system based on present huge square, and present huge square is the field pattern coding, and then data acess method is the access of field pattern mode during filtering, and anti-is as the same.Only under the last special status of chatting, though present huge square is the hardwood graph code, when the huge square with the top field pattern carried out filtering, data must adopt the access of field pattern mode, by the time finish and the huge square in top between two boundary filterings after, just switch back with the access of hardwood figure mode.
Seeing also Fig. 7, is that present huge square is that hardwood graph code, the huge square in top are the filter sequence schematic diagram of field pattern coding in the block ripple device for the present invention removes; And Fig. 8 removes filter sequence schematic diagram under the block ripple device normal condition for the present invention.In design,, the order of filtering is changed over Fig. 7 and shown in Figure 8 for various boundary condition.Among Fig. 7 when present huge square be the hardwood graph code, above huge square be field pattern when coding, article one border that do left side vertical direction must use the pixel in two block of cells in top.This is because of being the field pattern coding when the top, when carrying out the filtering action, in must huge square by present hardwood graph code, taking out four all is the pixel of field, top or field, the end, so the filtering time of vertical direction must postpone backward, after finishing horizontal direction filtering, last two block of cells could begin to carry out.And Fig. 8 is the filter sequence schematic diagram under the deblocking filter normal condition of the present invention, same observing done horizontal direction earlier, just do the rule of vertical direction and do filtering, but the direction of filtering is article one border by left side vertical direction to begin to make the direction that general filtering different from the past is carried out.The working area of median when secondly also having had filtering concurrently because of output buffer 106, the therefore action of with the time same address being read and write for fear of first temporary storage buffer region 1061 is so changed the direction that filtering is carried out.
In order further to explain the direction that filtering is carried out and the filtering action of huge square, then seeing also Fig. 4 deblocking filter of the present invention is the system Organization Chart, and Fig. 7 the present invention removes, and present huge square is that hardwood graph code, the huge square in top are the filter sequence schematic diagram of field pattern coding in the block ripple device.Present huge square is the hardwood graph code in deblocking filter of the present invention, the top is huge and square is the following time of filtering situation of field pattern coding, is to do the huge square of top brightness earlier then to do following two huge squares of colourity equally.Its filtering action is that horizontal filter 103 is the horizontal filterings (numbering 1,2,3,4 as Fig. 7) that in regular turn block of cells of the huge square vertical direction of brightness carried out first limit at the beginning, and behind the horizontal filtering that executes first limit, writes in the output buffer 106.Horizontal filter 103 is read the block of cells of brightness square vertical direction in regular turn in output buffer 106 then, and carry out the horizontal filtering (numbering 5,6,7,8) on second limit as Fig. 7, and behind the horizontal filtering that executes second limit, write in the transpose buffering district 105.And because of when the top being the field pattern coding, when carrying out the filtering action, in must huge square by present hardwood graph code, taking out four all is the pixel of field, top or field, the end, so the filtering time (numbering 8 as Fig. 7) of block of cells A vertical direction must postpone backward, so that the vertical filtering of block of cells A is that by the time block of cells A and block of cells B (block of cells to) begin to carry out after finishing the filtering of horizontal direction again.And in like manner in block of cells E, I, M, and block of cells Q, S, U, the W of colourity square.When carrying out the vertical filtering of block of cells A, be in transpose buffering district 105, to read the block of cells A (finishing the block of cells A of filtering numbering 1,5 as Fig. 7) that finishes first limit and the second limit horizontal filtering, with the first limit vertical filtering (carrying out the filtering of block of cells A numbering 8 as Fig. 7) of carrying out block of cells A, and then block of cells A is temporary in the transpose buffering district 105.In like manner in transpose buffering district 105, read the block of cells B (finishing the block of cells B of numbering 2,6 filtering as Fig. 7) that finishes first limit and the second limit horizontal filtering then, to carry out the vertical filtering of block of cells B numbering 9.The block of cells of finishing level and vertical filtering is temporary in output buffer 106 (finishing the block of cells A of numbering 1,5,8,9 filtering as Fig. 7).Just huge square is write back in the DRAM (Dynamic Random Access Memory) 2 by output buffer 106 with full luminance after all block of cells in the huge square of brightness (block A ~ block P) are all finished filtering at last.And in like manner carry out the filtering of the huge square of colourity.
Then seeing also Fig. 4 deblocking filter of the present invention is the system Organization Chart, and the filter sequence schematic diagram under Fig. 8 deblocking filter normal condition of the present invention.When the filtering of (pure hardwood figure or pure field pattern) under the deblocking filter normal condition of the present invention, be to do the huge square of top brightness earlier then to do following two huge squares of colourity.And horizontal filter 103 is the horizontal filterings (numbering 1,2,3,4 as Fig. 8) that in regular turn block of cells of the huge square vertical direction of brightness carried out first limit at the beginning, and behind the horizontal filtering that executes first limit, writes in the output buffer 106.Horizontal filter 103 is read the block of cells of brightness square vertical direction in regular turn in output buffer 106 then, and carry out the horizontal filtering (numbering 5,6,7,8) on second limit as Fig. 8, and behind the horizontal filtering that executes second limit, write in the transpose buffering district 105.When horizontal filtering is carried out numbering 7 the time, in transpose buffering district 105, read the block of cells A (finishing the block of cells A of numbering 1,5 filtering as Fig. 8) that finishes first limit and the second limit horizontal filtering, with the first limit vertical filtering (carrying out the filtering of block of cells A numbering 7 as Fig. 8) of carrying out block of cells A, and then block of cells A is temporary in the transpose buffering district 105.In like manner do the vertical filtering of block of cells B numbering 8 then.This block of cells of finishing level and vertical filtering is temporary in output buffer 106 (finishing the block of cells A of numbering 1,5,7,8 filtering as Fig. 8).Just huge square is write back in the DRAM (Dynamic Random Access Memory) 2 by output buffer 106 with full luminance after all block of cells in the huge square of brightness (block A ~ block P) are all finished filtering at last.And in like manner carry out the filtering of the huge square of colourity.
But, the above, detailed description and accompanying drawing for the preferable specific embodiment of the present invention, feature of the present invention is not limited thereto, and is not that all scopes of the present invention should be as the criterion with claim in order to restriction the present invention, all closing in the embodiment of the spirit variation similar of claim of the present invention with it, all should be contained in the category of the present invention, anyly be familiar with this operator in the field of the invention, can think easily and variation or modify all can be encompassed within the claim category of this case.
Claims (18)
1. a deblocking filter is characterized in that, comprising:
One horizontal filter is to be used to receive a huge square, and this huge square is carried out the filtering of horizontal direction;
One vertical filter is to be used for this huge square is carried out the filtering of vertical direction;
One output buffer is to be connected between this horizontal filter and this vertical filter, is used for temporary through filtered this huge square of horizontal direction and through filtered this huge square of vertical direction;
One transpose buffering district, be to be connected in this horizontal filter, this vertical filter and this output buffer, receive the block of cells of having finished filtered this huge square of horizontal direction, the block of cells transposition that is used for finishing filtered this huge square of horizontal direction carries out the filtering of vertical direction for this vertical filter;
One Data Management Unit is to be connected in output state, in order to judge the encoding state of this huge square, reaches this huge square that horizontal filtering and vertical filtering are finished in output; And
One control unit, be to be connected in this horizontal filter, this vertical filter, this transpose buffering district and this Data Management Unit, in order to control this horizontal filter, this vertical filter, this transpose buffering district and the execution of this Data Management Unit filtering action to this huge square.
2. deblocking filter as claimed in claim 1 is characterized in that this huge square is to be huge square of a colourity or the huge square of a brightness.
3. deblocking filter as claimed in claim 1 is characterized in that comprising an input buffer, is to be connected in this horizontal filter, be used to store will carry out filtering this huge square so that this horizontal filter filtering to be provided.
4. deblocking filter as claimed in claim 3 is characterized in that this huge square is to comprise huge square of a colourity and the huge square of a brightness.
5. deblocking filter as claimed in claim 1 is characterized in that comprising a computing comparing unit, is acquisition and the comparison that is used for this huge square parameter, judges for this control unit whether this present huge square needs to carry out filtering.
6. deblocking filter as claimed in claim 1 is characterized in that comprising a data processing unit, is to be connected in this information management unit, is used to detect adjacent and the encoding state of an adjacent huge square of this huge square at present.
7. deblocking filter as claimed in claim 1, it is characterized in that comprising a direct memory access unit, be to be connected between this Data Management Unit and a DRAM (Dynamic Random Access Memory), be used to produce the address of this lively attitude random access memory, for this Data Management Unit with this huge square of finishing filtering in this output buffer this DRAM (Dynamic Random Access Memory) of writing direct.
8. deblocking filter as claimed in claim 1 is characterized in that this output buffer is to comprise:
One first temporary storage buffer region is to be used for storing this huge square;
One second temporary storage buffer region is to be used to store an adjacent huge square required when carrying out this huge square Far Left boundary filtering; And
One buffer array is to be used for storing carrying out this huge square topmost during boundary filtering, the required huge square of using in a top.
9. deblocking filter as claimed in claim 8 is characterized in that this second temporary storage buffer region is the buffer for one 8 * 16 * 8.
10. deblocking filter as claimed in claim 8 is characterized in that this buffer array all is to carry out from same group of buffer in the input and output of all directions.
11. deblocking filter as claimed in claim 1 is characterized in that this transpose buffering district comprises many group buffers, with the block of cells of temporary this huge square.
12. deblocking filter as claimed in claim 11 is characterized in that those buffers are respectively is one 4 * 4 * 8 buffer.
13. one kind is removed the blocking filtering method, it is characterized in that step comprises:
The required huge square in a top when reading one first huge square filtering;
Carry out the filtering of one first huge square, read the required huge square in top of one second huge square simultaneously;
Carry out the filtering of one second huge square, the result with this first huge square filtering writes back in the DRAM (Dynamic Random Access Memory) simultaneously; And
The filtering result of this second huge square is write back in this DRAM (Dynamic Random Access Memory).
14. the blocking filtering method of removing as claimed in claim 13 is characterized in that the first huge square is to be the huge square of a brightness, the second huge square is to be the huge square of a colourity.
15. the blocking filtering method of removing as claimed in claim 14 is characterized in that the filter step of this first huge square normal condition comprises:
In regular turn a block of cells of this first huge square vertical direction is carried out the horizontal filtering on first limit, and behind the horizontal filtering that executes first limit, write in the output state;
In regular turn this block of cells of this first huge square vertical direction is carried out the horizontal filtering on second limit, and behind the horizontal filtering that executes second limit, write in the transposition buffer;
In this transposition working area, read this block of cells of finishing first limit and the second limit horizontal filtering, finish the first limit vertical filtering of this block of cells of first limit and the second limit horizontal filtering with execution;
Be temporary in this transposition buffer with this block of cells of finishing the first limit vertical filtering finishing first limit and the second limit horizontal filtering;
In this output state, read next block of cells of finishing first limit and the second limit horizontal filtering, to carry out the first limit vertical filtering of this next block of cells;
This block of cells of finishing level and vertical filtering is temporary in this output working area; And
After all block of cells are all finished filtering in this first huge square, just write back in this DRAM (Dynamic Random Access Memory).
16. the blocking filtering method of removing as claimed in claim 15 is characterized in that the filter step of this second huge square normal condition comprises:
In regular turn a block of cells of this second huge square vertical direction is carried out the horizontal filtering on first limit, and behind the horizontal filtering that executes first limit, write in the output state;
In regular turn this block of cells of this second huge square vertical direction is carried out the horizontal filtering on second limit, and behind the horizontal filtering that executes second limit, write in the transposition buffer;
In this transposition working area, read this block of cells of finishing first limit and the second limit horizontal filtering, finish the first limit vertical filtering of this block of cells of first limit and the second limit horizontal filtering with execution;
Be temporary in this transposition buffer with this block of cells of finishing the first limit vertical filtering finishing first limit and the second limit horizontal filtering;
In this output state, read next block of cells of finishing first limit and the second limit horizontal filtering, to carry out the first limit vertical filtering of this next block of cells;
This block of cells of finishing level and vertical filtering is temporary in this output working area; And
After all block of cells are all finished filtering in this second huge square, just write back in this DRAM (Dynamic Random Access Memory).
17. the blocking filtering method of removing as claimed in claim 14, it is characterized in that this first huge square is that hardwood graph code, the huge square in a top are the filtering of field pattern coding, the filtering time of the top block of cells vertical direction of this first huge square must postpone backward, and the vertical filtering that makes the top block of cells of this first huge square is to wait until after this top block of cells and this next block of top block of cells are all finished the filtering of horizontal direction to begin to carry out again.
18. the blocking filtering method of removing as claimed in claim 17, it is characterized in that this second huge square is that hardwood graph code, the huge square in this top are the filtering of field pattern coding, the filtering time of the top block of cells vertical direction of this second huge square must postpone backward, and the vertical filtering that makes the top block of cells of this first huge square is to wait until after this top block of cells and this next block of top block of cells are all finished the filtering of horizontal direction to begin to carry out again.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2006101387568A CN101184227A (en) | 2006-11-13 | 2006-11-13 | Deblocking filter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2006101387568A CN101184227A (en) | 2006-11-13 | 2006-11-13 | Deblocking filter |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101184227A true CN101184227A (en) | 2008-05-21 |
Family
ID=39449277
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2006101387568A Pending CN101184227A (en) | 2006-11-13 | 2006-11-13 | Deblocking filter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101184227A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102055971A (en) * | 2009-10-29 | 2011-05-11 | 财团法人工业技术研究院 | Deblocking filtering apparatus and method for video compression |
| CN102215324A (en) * | 2010-04-08 | 2011-10-12 | 安凯(广州)微电子技术有限公司 | Filtering circuit for performing filtering operation on video image and filtering method thereof |
-
2006
- 2006-11-13 CN CNA2006101387568A patent/CN101184227A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102055971A (en) * | 2009-10-29 | 2011-05-11 | 财团法人工业技术研究院 | Deblocking filtering apparatus and method for video compression |
| CN102055971B (en) * | 2009-10-29 | 2012-08-29 | 财团法人工业技术研究院 | Deblocking filtering apparatus and method for video compression |
| CN102215324A (en) * | 2010-04-08 | 2011-10-12 | 安凯(广州)微电子技术有限公司 | Filtering circuit for performing filtering operation on video image and filtering method thereof |
| CN102215324B (en) * | 2010-04-08 | 2013-07-31 | 安凯(广州)微电子技术有限公司 | Filtering circuit for performing filtering operation on video image and filtering method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101472173B (en) | Method, system and filter for filtering de-block | |
| CN1156154C (en) | Method for specifying image element to memory sequence inorder to compression encoding | |
| US20060133504A1 (en) | Deblocking filters for performing horizontal and vertical filtering of video data simultaneously and methods of operating the same | |
| US20080285652A1 (en) | Apparatus and methods for optimization of image and motion picture memory access | |
| US20050190976A1 (en) | Moving image encoding apparatus and moving image processing apparatus | |
| US20090129478A1 (en) | Deblocking filter | |
| JP2010527194A (en) | Dynamic motion vector analysis method | |
| JPH08123953A (en) | Picture processor | |
| US20040100472A1 (en) | Method and/or apparatus for video data storage | |
| CN110337002B (en) | HEVC (high efficiency video coding) multi-level parallel decoding method on multi-core processor platform | |
| CN1652605B (en) | Video codecs, data processing systems and methods for the same | |
| CN101123725A (en) | Hardware implementation method for H.264 block elimination effect filter | |
| CN1192106A (en) | Memory manager for MPEG decoder | |
| CN101635849B (en) | Loop filtering method and loop filter | |
| US20050259744A1 (en) | Video deblocking memory utilization | |
| CN104253998A (en) | Hardware on-chip storage method of deblocking effect filter applying to HEVC (High Efficiency Video Coding) standard | |
| CN101841722B (en) | Detection method of detection device of filtering boundary strength | |
| CN101184227A (en) | Deblocking filter | |
| US7536487B1 (en) | Low power memory hierarchy for high performance video processor | |
| KR101615466B1 (en) | Capturing multiple video channels for video analytics and encoding | |
| CN101304528B (en) | Method for mapping video processor video data and memory space | |
| EP1992162B1 (en) | Memory organizational scheme and controller architecture for image and video processing | |
| CN101459839A (en) | Deblocking effect filtering method and apparatus for implementing the method | |
| CN101005619B (en) | Loop circuit filtering method | |
| CN101223789A (en) | Image encoding device and image encoding method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20080521 |