CN1722842A - The filtering method of audio-visual codec and filter apparatus - Google Patents
The filtering method of audio-visual codec and filter apparatus Download PDFInfo
- Publication number
- CN1722842A CN1722842A CNA2005100922132A CN200510092213A CN1722842A CN 1722842 A CN1722842 A CN 1722842A CN A2005100922132 A CNA2005100922132 A CN A2005100922132A CN 200510092213 A CN200510092213 A CN 200510092213A CN 1722842 A CN1722842 A CN 1722842A
- Authority
- CN
- China
- Prior art keywords
- filtering
- macro block
- filter
- filter field
- carried out
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
- H04N19/86—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression involving reduction of coding artifacts, e.g. of blockiness
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/117—Filters, e.g. for pre-processing or post-processing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
- H04N19/423—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation characterised by memory arrangements
Abstract
About audio-visual codec, provide a kind of filtering method and filter apparatus of audio-visual codec.The filtering method of described audio-visual codec comprises: determine with the macro block to be filtering area in units; The presumptive area that further will be defined as on the borderline region of the macro block of filter field and last adjacent macroblocks covers in this filter field; And go into piece filtering according to what time sequencing had that causality ground carries out filter field.Therefore, by revising the non-causality loop filter that VC-1 adopted and predetermined buffer further being set, can be effectively to becoming blocking artefacts to carry out filtering and not increasing memory bandwidth or cause frame delay.
Description
The cross reference of related application
The application requires respectively on June 22nd, 2004 and on July 19th, 2004 benefit of priority the korean patent application of Korea S Department of Intellectual Property submission 10-2004-0046653 number and 10-2004-0055893 number, and the disclosure of these applications all is incorporated in this for your guidance.
Technical field
The present invention relates to a kind of audio-visual codec, and relate in particular to a kind of filtering method and filter apparatus of audio-visual codec.
Background technology
Recently, Microsoft is to having submitted the relevant information with VC-1 (being called VC9 in the past) to as the film and television Ssociety of engineers (SMPTE) of one of international standardizing bodies, and described VC-1 is a moving image compression standard draft.At present, examining the VC-1 draft standard energetically, and expecting that it is adopted as international standard the most at last.
All adopted as moving image compression standard and the Motion Picture Experts Group (MPEG)-2 that is widely used, MPEG-4 with H.264 the standard except each, expected that also VC-1 becomes leading moving image compression standard.Since its compression efficiency is approximate possess optimal compression efficiency compression efficiency H.264 80%, and its implementation complexity is equivalent to implementation complexity about 60% H.264, so VC-1 is considered to have good complexity performance than (complexity-to-performance ratio).Equally, with regard to picture quality, VC-1 also is considered to possess than MPEG-2 or the better quality of MPEG-4.
According to above-mentioned moving image compression standard, be that unit comes digital picture is carried out Code And Decode with the piece.At this moment, in the digital picture of reproducing by decode procedure, a kind of like this phenomenon takes place, that is: the border of piece looks like different.This phenomenon is called into blocking artefacts (blocking artifact), and in order to eliminate into blocking artefacts, uses loop filter.That is, loop filter is so that eliminate the filter that (removing into piece) becomes the piece phenomenon in a kind of coding loop that is inserted into codec (encoder).
Hereinafter, the decode procedure of VC-1 will be explained in more detail.
Fig. 1 is the block diagram of vC-1 decoder.With reference to Fig. 1, described VC-1 decoder comprises bit stream resolution unit, contrary Variable Length Code (vLC) unit, inverse quantization unit, inverse transformation block, motion compensation units, overlapping level and smooth and loop filtering unit, intensity compensation and scope unit for scaling and memory.Specifically, one embodiment of the present of invention relate to the loop filter in the element of decoder.
If the LOOPFILTER sequential parameter is set to 1 in the vC-1 decoder, then the reconstructed image of handling by contrary vLC unit, inverse quantization unit and inverse transformation block is carried out filtering.Should before being used as the reference frame that is used for motion compensation, reconstructed image apply this filtering.So, owing to after filtering, reconstructed image usefulness is acted on the reference frame of motion compensation, thereby decoder should strictly be followed the filtering by standard code.Because piece image is normally formed by a plurality of sections (slice), thereby each is cut into slices carry out loop filtering independently.
The main purpose of loop filtering is exactly to reduce or eliminate the one-tenth piece phenomenon that produces on block boundary because of quantification.At this moment, under the situation of I image, the piece with 8 * 8 is that unit carries out conversion, and thus, is unit level ground and vertically carries out filtering with 8 pixels.Just, carry out filtering on each in 8 pixels, 16 pixels and 24 pixels.
Simultaneously, under the situation of P image, depending on alternative types is 8 * 8,8 * 4 or 4 * 8, and block boundary may be vertically or flatly appeared on each 4 pixel cell.Thus, on each 4 pixel cell, vertically and flatly carry out filtering.Just, carry out filtering on each in 4 pixels, 8 pixels and 12 pixels.To more specifically make an explanation now to this.
1) loop filtering in the I image
In the I image, whole 8 * 9 block boundaries execution are gone into piece filtering to eliminate into blocking artefacts.Fig. 2 A and 2B are the figure that is illustrated in the pixel in the filtered horizontal/vertical block boundary in the I image.
With reference to Fig. 2 A and 2B, show the left-hand side part of the image that forms by element Y, Cb, Cr.Cross (+) remarked pixel, and circular (o) represents the pixel of filtering.Shown in Fig. 2 A and 2B, first vertical line to first horizontal line on the top and left-hand side does not carry out filtering.Equally, not to be positioned at the bottom on first horizontal line and the last vertical line (not shown) on the right-hand side carry out filtering.That is to say, suppose in a frame that N represents the quantity of 8 * 8 piece in the horizontal direction, and M represents the quantity of 8 * 8 piece in vertical direction, in the horizontal direction only to horizontal line (7,8), (15,16) ... ((N-1) * 8, (N-1) * 8) carries out filtering.Equally, in vertical direction only to vertical line (7,8), (15,16) ... ((M-1) * 8, (M-1) * 8) carries out filtering.
In addition, the order of filtering also is very important.That is to say, in a complete frame, only after all horizontal sides boundary lines are all filtered, just filtering is carried out in all vertical edges boundary lines.So, for a frame is carried out filtering, filtering is carried out in all boundary lines on the horizontal direction, and the result is stored in the memory, then filtering is carried out in all boundary lines on the vertical direction.Because this non-causality filtering method (non-causal filtering method), one frame is carried out filtering need big bandwidth, and since should finish in the horizontal direction with this both direction of vertical direction on the filtering carried out so that finish filtering, thereby frame delay may appear.
2) loop filtering in the P image
In the P image, piece may be under the internal schema (intra mode) or under the middle model (inter mode).Because the piece that is under the internal schema always adopts 8 * 8 conversion, therefore only 8 * 8 block boundary is carried out filtering.Simultaneously, the piece under the pattern that mediates adopts 8 * 8,8 * 4,4 * 8 or 4 * 4 inverse transformation, so that residual error is decoded.So,, can carry out filtering or it not carried out filtering the border between current block and the adjacent block according to the situation of adjacent block.Judge carry out filtering to which block boundary (8 * 8) or sub-block boundary (8 * 4,4 * 8 or 4 * 4) according to following standard.
(1) unconditionally filtering is carried out on the border of the coded sub-blocks in 8 * 8 the piece (8 * 4,4 * 8 or 4 * 4).
(2) if block (or sub-piece) has identical motion vector with adjacent block (or sub-piece) and the both does not have residual error, then described block boundary is not carried out filtering.In addition, other block boundary carries out filtering to all.
Fig. 3 is the figure that is illustrated in the example of block boundary filtered in the P image.With reference to Fig. 3, show the filtering of in the border of adjacent block, carrying out.The motion vector of supposing piece is identical (if inequality, then unconditionally carrying out filtering), is the encoding block with conversion coefficient with the piece (or sub-piece) of grey colour specification.The piece (or sub-piece) of white is the piece with any conversion coefficient.Thick line is represented the border of filtering, and fine rule is represented unfiltered border.According to two above-mentioned standards, determine to want the filtered piece or the border of sub-piece.Although the example among Fig. 3 only shows two situations that the piece level is adjacent, also same procedure can be applied to the vertical adjacent situation of two pieces.
Fig. 4 A and 4B are the figure that is illustrated in the piece level and/or the pixel in the vertical boundary of filtering in the P image.With reference to Fig. 4 A, show the filtered horizontal boundary of energy.
Owing to can carry out filtering to 8 * 8 and 8 * 4 horizontal boundary, thus each the the the 4th and the 5th, the 8th and the 9th and the 12 and the tenth three-way can be can be filtered boundary pixel.
Simultaneously, with reference to Fig. 4 B, show the filtered vertical boundary of energy.Owing to can carry out filtering to 8 * 8 and 4 * 8 vertical boundary, thus each the the the 4th and the 5th, the 8th and the 9th and the 12 and the tenth three-way can be can be filtered boundary pixel.Be similar to the I image, not to first and last vertical line and horizontal line carry out filtering.
In addition, same under the situation of P image, filter sequence also is very important.At first, for entire frame, begin continuously all horizontal boundaries of 8 * 8 to be carried out filtering from top line.The filtering result of 8 * 8 horizontal boundaries is stored in the memory.Secondly, for entire frame, begin continuously all horizontal boundaries of 8 * 4 to be carried out filtering from top line equally.Filtering is the result equally also be stored in the memory.The 3rd, for the entire frame on the vertical direction, begin all vertical boundaries of 8 * 8 are carried out filtering from left-hand side, and the result is stored in the memory.At last, for entire frame, begin all block boundaries of 4 * 8 are carried out filtering from left-hand side equally, and the result is stored in the memory.That is,, come all horizontal boundaries are carried out filtering, come all vertical boundaries are carried out filtering by two steps then by two steps for an entire frame.Thus, because the non-causality filtering method for a frame is carried out filtering, should be carried out four steps of filtering, so frame delay may occur.
3) loop filtering in the B image
The filtering of B image is identical with the filtering of above-mentioned I image.To explain the process of in the border of each piece (or sub-piece), carrying out filtering now in more detail.Fig. 5 A and 5B are the figure that is illustrated in the example of filtered pixel in the border of adjacent block.
No matter vertical direction or horizontal direction, the minimum number of contiguous pixels that should be filtered is 4, and the quantity of the whole pixels in vertical line or the horizontal line is 4 multiple.So, can be that unit carries out filtering with 4 pixels.With reference to Fig. 5 A, for example,, then each 8 pixel is divided into the section of two 4 pixels if the border of two pieces is to be formed by a pair of 8 pixels.In the section of each 4 pixel, at first the 3rd pair of pixel by the X mark as shown in the figure carried out filtering.Equally, according to the result of filtering, determine whether to carry out to of the filtering of its excess-three to pixel.Do not go out ready-made if the 3rd pair of comparative result by the pixel of X mark shows, therefore do not need to carry out filtering, so just can omit of the filtering of its excess-three to pixel.
With reference to Fig. 5 B, show the filtering of in the 3rd pair of pixel, carrying out.Pixel P4 and P5 are the pixels that can be exchanged in filtering.Acquisition has been used from pixel P4 and P5 to it and has been begun pixel (P1 to P4) on the left-hand side and predetermined backoff value weights, mean value between the pixel (P5 to P8) on the right-hand side, and, just can eliminate or reduce the appearance of the sudden change in the pixel value in the border of piece by described value being added on each of pixel P4 and P5.In this case, shown in above-mentioned Fig. 5 A,, so equally also carry out, otherwise omit, handle the section of next 4 pixels then of the filtering of its excess-three to pixel to of the filtering of its excess-three to pixel if the 3rd pair of pixel need be carried out filtering.
As mentioned above, in the filtering of the one-tenth blocking artefacts of eliminating VC-1, be that filtering is carried out on unit level ground with frame or section, and vertically carry out filtering subsequently, therefore the result of previous horizontal filtering should be stored in the memory.So have such problem: the memory bandwidth that needs are big unnecessary frame delay occurs, and frame memory increases.Because therefore this non-causality of VC loop filtering is difficult to realize the hardware compression device.
Summary of the invention
The invention provides a kind of filtering method and a kind of filter apparatus, wherein utilize described method, carry out filtering effectively and do not increase memory or bandwidth by the non-causality loop filtering that improves in the audio-visual codec.
According to an aspect of the present invention, provide a kind of filtering method of audio-visual codec, having comprised: determined with the macro block to be filtering area in units; Further the presumptive area on the borderline region of macro block and last adjacent macroblocks is covered in the filter field; And according to time sequencing, have causality ground (causally) carry out to the filter field after augmenting go into piece filtering, the filter field after described the augmenting further comprises the presumptive area of the borderline region between the macro block.
Presumptive area on the borderline region of described macro block can comprise a plurality of pixels with minimum 4 pixel units.
Described to go into piece filtering be by using minimum 4 pixel values on each direction from the both direction that the borderline region of macro block begins to carry out, and with respect to the borderline region as the macro block of center line pixel value is set symmetrically.
Describedly go into piece filtering and can carry out, utilize described method, in the horizontal direction a filter field is carried out filtering, carry out filtering then in vertical direction, then next filter field is carried out filtering according to the causality method.
Describedly go into piece filtering and can be performed, make in the horizontal direction 8 * 8 block boundary is partly carried out filtering, then 8 * 4 block boundary is partly carried out filtering, and in vertical direction 8 * 8 block boundary is partly carried out filtering, then 4 * 8 block boundary is partly carried out filtering.
Described audio-visual codec meets the VC-1 standard of film and television Ssociety of engineers (SMPTE).
According to a further aspect in the invention, provide a kind of filter apparatus of audio-visual codec, having comprised: buffer unit, it is unit storage filter field with the macro block; Line buffer and column buffer, the presumptive area of the boundary member of macro block of being stored in its storage buffer unit and the previous macro block adjacent with this macro block of storing; And filter unit, it is according to time sequencing, has causality ground to carry out the presumptive area that is stored in filter field in buffer unit, line buffer and the column buffer and boundary member gone into piece filtering.
The part of other aspects of the present invention and/or advantage will be set forth in explanation subsequently, and a part can clearly learn from described explanation, perhaps know by putting into practice the present invention.
Description of drawings
By below in conjunction with the explanation of accompanying drawing to embodiment, these and/or others of the present invention and advantage will become clear and be more readily understood, in described accompanying drawing:
Fig. 1 is the block diagram of VC-1 decoder;
Fig. 2 A and 2B are the figure that is illustrated in the pixel in the horizontal/vertical block boundary filtered in the I image;
Fig. 3 is the figure that is illustrated in the example of block boundary filtered in the P image;
Fig. 4 A and 4B are the figure that is illustrated in piece level and/or the pixel in the vertical boundary filtered in the P image;
Fig. 5 A and 5B are the figure that is illustrated in the example of pixel filtered in the border of adjacent block;
Fig. 6 is the block diagram of filter apparatus according to an embodiment of the invention;
Fig. 7 is the figure that illustrates according to an embodiment of the invention, will carry out it frame of filtering;
Fig. 8 be illustrate according to an embodiment of the invention, in the 1st kind of situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it;
Fig. 9 be illustrate according to an embodiment of the invention, in the 2nd kind of situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it;
Figure 10 A and 10B be illustrate according to an embodiment of the invention, in the 3rd kind of situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it;
Figure 11 A and 11B be illustrate according to an embodiment of the invention, in the 4th kind of situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it; With
Figure 12 be illustrate according to alternative embodiment of the present invention, after level and vertical division be that unit carries out the figure of the filter sequence of filtering by it with the piece.
Embodiment
Now will be in detail with reference to embodiments of the invention, illustrate example of the present invention in the accompanying drawings, wherein identical Reference numeral refers to components identical all the time.These embodiment are described below with reference to the accompanying drawings so that explain the present invention.
Fig. 6 is the block diagram of filter apparatus according to an embodiment of the invention.With reference to Fig. 6, with the frame be unit carry out filtering conventional vC-1 filter by comparison, filter apparatus is characterised in that according to an embodiment of the invention: be that unit carries out filtering with the macro block.For this reason, further be provided with and be used for interim storage about the independent line buffer 630 of the Pixel Information of the boundary member of last macro block and independent column buffer 640.
Just, filter apparatus has according to an embodiment of the invention: line buffer (ROW_BUF) 630 and the column buffer (COL_BUF) 640 that are used to eliminate into the filter unit 610 of blocking artefacts, the buffer unit (L_BUF) 620 that is used to store the data of wanting filtered and filtering result and the Pixel Information of the boundary member that is used to store relevant last macro block.
The basic processing unit of VC-1 codec is the macro block of 16 * 16 pixel sizes.Therefore, be that unit carries out loop filtering also with 16 * 16 macro block.Buffer unit 620 has the size of 16 * 16 pixels, and stores intermediate object program and the final result that 16 * 16 macro block is carried out filtering.
In an embodiment of the present invention, specify filter field for macro block.Then, filter field is augmented and filtering.Augmenting operation can comprise: the pixel in the filter field adjacent with the borderline region of this macro block.With this border adjacent pixels comprise at least boundary pixel and with the boundary pixel adjacent pixels, and describe as Fig. 6, can comprise triplex row/row and boundary pixel adjacent pixels.In addition or alternatively, augment operation and can comprise last time filtered macro block in the filter field.
In order to explain the operation of filter apparatus according to an embodiment of the invention, suppose to exist in the horizontal direction 4 macroblock size and a frame of 3 macroblock size in vertical direction.Fig. 7 illustrates according to an embodiment of the invention, will carry out the figure of the frame of filtering to it.
With reference to Fig. 7, each macro block all has numbering.In order to be that unit carries out filtering according to embodiments of the invention, will explain 4 kinds of situations of macro block filtering operation now with the macro block.Specifically, will be explained on the horizontal direction, in as first kind of situation zone of the filter field of macro block MB1 and the filtering operation that in second kind of situation zone, carries out as the filter field of macro block MB2.Equally, will be explained on the vertical direction, in as the third situation zone of the filter field of macro block MB5 and the filtering operation that in the 4th kind of situation zone, carries out as the filter field of macro block MB6.
Fig. 8 be illustrate according to an embodiment of the invention, in the 1st kind of situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it.
If the motion compensation of a macro block (with reference to Fig. 1) finishes, then carry out the loop filtering of macro block.First kind of situation shows the loop filtering process of macro block MB1.
With reference to Fig. 6 and 8, the macro block MB1 after the motion compensation is stored in the buffer unit (L_BUF) 620.Depending on is any situation, and macro block MB1 that will its overlapping conversion has finished after motion compensation is stored in the buffer unit (L_BUF) 620.Fig. 8 (A) shows the filtering target area of first kind of situation.Because macro block MB1 is first macro block of first section, so line buffer 630 and column buffer 640 all be sky.
With reference to Fig. 8 (B), show the filtering of the block boundary of horizontal direction 8 * 8.Pixel on the block boundary of the horizontal direction 8 * 8 among the macro block MB1 that is labeled as the black circle is carried out filtering.When macro block MB5 is filtered, boundary member below the MB1, macro block MB1 and macro block MB5 is carried out filtering.Equally, when adjacent macroblocks MB2 is filtered, 4 pixel columns on the right-hand side of macro block MB1 are carried out filtering.Therefore, the pixel on the part of being represented by thick dashed line is the pixel that is used for the black circular pixel is carried out filtering.
In the part of representing by thick dashed line, will be input in the p0 shown in Fig. 6, p1, p2, p3, q0, q1, q2, the q3 latch along continuous as the column vector of all 12 vertical direction pixel row.Every row all have 8 pixels.That is to say, as shown in the figure, 4 top pixels are input among p3, p2, p1 and the p0, and 4 following pixels are input among q0, q1, q2 and the q3.
Equally, at the top of Fig. 8 (B), show the input sequence of 12 column vectors.
Just, as shown in Figure 5, owing at first the 3rd pair of pixel in the section of 4 pixels handled, at first in the section of each 4 pixel, first, the 5th and the 9th column vector carried out filtering.According to the filtering result, determine whether the rest of pixels in the section of 4 pixels carrying out filtering.The pixel that is input among p3, p2, p1, p0, q0, q1, q2 and the q3 is input to the filter unit 610 shown in Fig. 6, and in one-period, it is carried out filtering.The process of carrying out actual loop filtering is as reference Fig. 5 is described.In the process of carrying out actual loop filtering, can use any filtering method that comprises by the filtering method of VC-1 suggestion.In the filtering result, only will be input among the P0 and Q0 of Fig. 6 with the corresponding pixel of the block boundary that is labeled as the black circle.When determining to carry out filtering, just, when definite block boundary is not the profile of image partly, but when going out the part of ready-made blocking artefacts, then use the analog value in the value update buffer unit (L_BUF) 620 among P0 and the Q0.
With the same manner, carry out filtering to the the the 5th, the 6th, the 7th and the 8th column vector, carry out then the 9th, the tenth, the 11 and the filtering of the 12 column vector.As mentioned above, although all column vectors that comprised in the target area are carried out filtering, actual the borderline pixel execute store to piece upgrades.For example, under the situation of I image, only on the border of 8 * 8 piece, carry out and upgrade, and under the situation of P image, only on the border between the encoding block, carry out and upgrade.
Simultaneously, with reference to Fig. 8 (C), show the filtering of the block boundary of vertical direction 8 * 4.To carrying out filtering by the pixel on the block boundary of the vertical direction 8 * 4 among the macro block MB1 of the circular mark of black.Actual filtering identical with described in Fig. 8 (B).
If the filtering to the pixel on the horizontal direction block boundary among the macro block MB1 finishes, carry out filtering so to the pixel on the vertical direction block boundary among the macro block MB1.At first, shown in Fig. 8 (D), the pixel on the block boundary of the vertical direction among the macro block MB1 8 * 8 is carried out filtering.At this moment, when macro block MB5 is filtered, carry out filtering to 8 the following pixel columns on the block boundary.Equally, when adjacent macroblocks MB2 is filtered, 4 pixel columns on the right-hand side of macro block MB1 are carried out filtering.Therefore, the pixel in the part of being represented by thick dashed line is to be used for the black circular pixel is carried out the pixel of filtering.Next, shown in Fig. 8 (E), the pixel on the block boundary of the vertical direction among the macro block MB1 4 * 8 is carried out filtering.
Although only changed vertical direction and horizontal direction, actual filtering is with identical with the described filtering of 8 (C) with reference to Fig. 8 (B).The filtering result is stored in the buffer unit 620 temporarily, and is recorded in the external frame memory.Under the situation of macro block MB1, owing to only changed pixel in the zone shown in Fig. 8 (A), thereby with the data record in the zone shown in Fig. 8 (A) externally in the frame memory.Fig. 8 (F) shows will be as the outcome record of the filtering content in the frame memory externally.Then, in order to carry out to filtering as the MB2 of next macro block, will with 8 * 16 part cargo shipments of pixel on the right-hand side of the corresponding macro block MB1 of the boundary member of two pieces in column buffer 640.16 * 16 part cargo shipments of the pixel of macro block MB2 that equally, will be filtered in buffer unit 620 (with reference to Fig. 6).
Simultaneously, Fig. 9 be illustrate according to the embodiment of the invention, in second kind of situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it.With reference to Fig. 9, show the filtering of macro block MB2.Fig. 9 (A) shows the filtering target area.Except that macro block MB2, when macro block MB1 was filtered, 8 * 16 parts that temporarily are stored in the pixel in the column buffer 640 were also contained in the target area.
At first, in Fig. 9 (B) with on the block boundary of horizontal direction 8 * 8, want filtered pixel be labeled as the black circle.When a section or next macro block are filtered instantly, 8 * 8 following parts of pixel and 4 pixel columns on the right-hand side are handled, and only be used to the pixel that is marked as the black circle is carried out filtering by the pixel in the part of thick dashed line mark.That is to say, 16 column vectors altogether are input in the p shown in Fig. 6 and q the latch continuously and in addition filtering that each of described 16 column vectors all is that the row by 8 pixels form.Concrete filtering with describe with reference to Fig. 8 identical.Next, shown in Fig. 9 (C), the filtering of the horizontal boundary of execution 8 * 4.Described filtering is same as described above.
Simultaneously, in vertical direction, 8 * 8 vertical boundary filtering and 4 * 8 vertical boundary filtering have been shown in Fig. 9 (D) and 9 (E) respectively.With reference to Fig. 9 (D), import 8 row vectors and in addition filtering in the zone that is illustrated by the broken lines continuously according to the order that illustrates on the side leftward.That is to say, respectively 8 pixels of each vector are input in p3, p2, p1, p0, q0, q1, q2 and the q3 latch and in addition filtering.Although only changed vertical direction and horizontal direction, come the section of each 4 pixel is carried out filtering in the mode identical with the described mode of reference Fig. 5.Next, shown in Fig. 9 (E), 4 * 8 vertical boundary pixels are carried out filtering.In addition, similarly, the filtering to the vertical boundary of all the other vertical boundaries of 8 * 8 and 4 * 8 has been shown in Fig. 9 (F) and 9 (G) respectively.Fig. 9 (H) shows as the filtering result will be recorded in content in the external frame memory.Then, for next macro block MB3 is carried out filtering, will with 8 * 16 part cargo shipments of pixel on the left-hand side of the corresponding macro block MB2 in the border of two pieces in column buffer 640.16 * 16 part cargo shipments of the pixel of macro block MB3 that in addition, will be filtered in buffer unit 620 (with reference to Fig. 6).Thus, execution is to the filtering of the macro block MB1 to MB4 on the horizontal direction in first section.
Figure 10 A and 10B be illustrate according to an embodiment of the invention, in the third situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it.With reference to Figure 10 A and 10B, show the filtering of the first macro block MB5 of second section.After macro block MB4 is carried out filtering, 16 * 8 following pixels of macro block MB1 are loaded in the line buffer 630.In addition, 16 * 16 data of macro block MB5 that will be filtered are loaded in the buffer unit 620.
At Figure 10 (B) filtering to the boundary pixel of the boundary pixel of the horizontal direction among the macro block MB5 8 * 8 and 8 * 4 has been shown to 10 (E).Although only changed object pixel, filtering principle with reference to Fig. 8 or 9 described identical.In addition, filtering to the boundary pixel of the boundary pixel of vertical direction 8 * 8 and 4 * 8 has been shown in Figure 10 (F) and 10 (G).In Figure 10 (H), the filtering result has been shown, and the end value of filtering has been recorded in the external memory storage.
Then, for next macro block being carried out filtering, the content of update buffer unit 620, line buffer 630 and column buffer 640.Should be noted that: at this moment, the pixel that be updated is not the pixel of the front before the filtering, but carries out the resulting pixel of filtering.
Figure 11 A and 11B be illustrate according to an embodiment of the invention, in the 4th kind of situation zone shown in Figure 7 to be unit with the piece carry out the figure of embodiment of the filter sequence of filtering by it.With reference to Figure 11 A and 11B, (B) show filtering, and (F) show filtering vertical direction block boundary pixel to (I) to horizontal direction block boundary pixel to (E).Pixel in the thick dashed line is recorded in the external frame memory.Carry out the filtering of all the other macro blocks according to identical principle.
Thus, and be that the conventional VC-1 filter that unit carries out filtering is compared with the frame, filter apparatus is characterised in that according to an embodiment of the invention: be that filtering is carried out in the unit with the macro block.For this reason, further comprise independent line buffer 630 and column buffer 640, so that the Pixel Information of the boundary member of the relevant last macro block of interim storage.
Just, be the filtering of the block boundary part of unit executive level 8 * 8 and 8 * 4 with the macro block, and carry out the filtering of the block boundary part of vertical direction 8 * 8 and 4 * 8.The pixel that will be used for to next macro block carries out on the boundary member of filtering is stored in line buffer or the column buffer, next macro block is carried out in the filtering so that described pixel energy is used in.
Therefore, make it possible to carry out causality filtering, wherein utilize described causality filtering to come to carry out filtering by the order of macro block according to time sequencing.As a result, and be that unit carries out filtering and compares with the frame, required memory bandwidth reduces, and can prevent frame delay.
The embodiment that describes with reference to Fig. 8 to 11B is to be the filtering method (method 1) that unit handles with the macro block, and described method is suitable for reducing memory bandwidth.Yet method 1 needs more complicated control.Will increase memory bandwidth a little with reference to the replacement method (method 2) that Figure 12 explains, but need more uncomplicated control with respect to method 1.
Figure 12 be illustrate according to alternative embodiment of the present invention, after level and vertical division be that unit carries out the figure of the filter sequence of filtering by it with the piece.With reference to Figure 12, in method 2, horizontal filtering and vertical filtering are separated.
Just, for macro block MB1, filtering is carried out in such as Fig. 8 (C) and (D), and for all the other macro blocks in the same slice, carries out filtering with the same manner.
For the macro block MB5 of next section, filtering is carried out in such as Figure 10 (B) and (C), and for the macro block that is comprised in second section, carries out filtering with the same manner.Identical to the filtering that all the other sections are carried out with the filtering that second section is carried out.Yet, in the end in section, also need to be used for 16 * 8 parts of bottom are carried out the process of filtering.
If the filtering that so is through with and carries out is in the horizontal direction then carried out filtering with the same manner in vertical direction.That is to say that it is different having only direction, and filtering method is identical.If finish filtering that each macro block is carried out, then should be externally in the frame memory with outcome record.
For the purpose of explaining, the size of buffer unit, line buffer and column buffer is restricted to preliminary dimension, but in the enforcement of reality, can freely determine this size.For example, to be expanded be the size of a section to the horizontal direction size maximum of buffer unit and line buffer.
In this case, the bandwidth that is loaded onto in the memory of the data of line buffer is reduced.Just, owing to the increase of memory bandwidth along with buffer size reduces, therefore can suitably carry out the adjusting between the two.
Filtering method and equipment in the VC-1 codec have been described above.Yet, described method and apparatus can be applied to comprise in many ways other standard of loop filter.Just, should only regard as described embodiment is purpose illustrative but not in order to limit, and those skilled in the art will appreciate that: do not breaking away under the situation of the spirit and scope of the present invention that limit as claim, can make the various changes on form and the details.
According to aforesaid embodiments of the invention, a kind of filtering method and filter apparatus thereof in audio-visual codec is provided, described method is that unit has causality ground to carry out filtering with the macro block according to time sequencing.That is to say, by revising the non-causality loop filter that VC-1 adopted and predetermined buffer further being set, can be effectively to becoming blocking artefacts to carry out filtering and not increasing memory bandwidth or cause frame delay.
Although illustrated and described several embodiments of the present invention, but one of ordinary skill in the art will recognize that: can make change to these embodiment without departing from the principles and spirit of the present invention, scope of the present invention is to limit in claim and equivalent thereof.
Claims (22)
1. the filtering method of an audio-visual codec comprises:
Determine with the macro block to be filtering area in units;
Further the presumptive area on the borderline region of macro block and last adjacent macroblocks is covered in this filter field; And
Go into piece filtering according to what time sequencing had that causality ground carries out filter field, described filter field further comprises the presumptive area of the borderline region between the macro block.
2. the method for claim 1, wherein the presumptive area on the borderline region of described macro block comprises a plurality of pixels with minimum 4 pixel units.
3. the method for claim 1, wherein, by using minimum 4 pixel values on each direction from the both direction that the borderline region of macro block begins to carry out the described piece filtering of going into, with respect to the pixel value of assigning to be provided with symmetrically as the boundary portion of the macro block of center line.
4. method as claimed in claim 3, wherein, carry out the described piece filtering of going into by the causality method, wherein utilize described causality method, in the horizontal direction a filter field is carried out filtering, in vertical direction it is carried out filtering then, then next filter field is carried out filtering.
5. method as claimed in claim 4, wherein, in going into piece filtering, in the horizontal direction, block boundary to 8 * 8 partly carries out filtering, then 8 * 4 block boundary is partly carried out filtering, and in vertical direction, block boundary to 8 * 8 partly carries out filtering, then 4 * 8 block boundary is partly carried out filtering.
6. the method for claim 1, wherein described audio-visual codec meets the VC-1 standard of film and television Ssociety of engineers (SMPTE).
7. the filter apparatus of an audio-visual codec comprises:
Buffer unit, it is unit storage filter field with the macro block;
Line buffer and column buffer, the presumptive area of the boundary member of macro block of being stored in its storage buffer unit and the last macro block adjacent with this macro block of storing; With
Filter unit, it has causality ground to carry out according to time sequencing the presumptive area that is stored in filter field in buffer unit, line buffer and the column buffer and boundary member gone into piece filtering.
8. equipment as claimed in claim 7, wherein, the presumptive area of the borderline region of at least one described macro block comprises a plurality of pixels with minimum 4 pixel units.
9. equipment as claimed in claim 7, wherein, by using minimum 4 pixel values on each direction from the both direction that the boundary member of macro block begins to carry out the described piece filtering of going into, with respect to the pixel value of assigning to be provided with symmetrically as the boundary portion of the macro block of center line.
10. equipment as claimed in claim 9, wherein, described filter unit is carried out filtering by the causality method, wherein utilize described causality method, in the horizontal direction a filter field is carried out filtering, in vertical direction it is carried out filtering then, then next filter field is carried out filtering.
11. equipment as claimed in claim 10, wherein, described filter unit is carried out filtering, so that in the horizontal direction, block boundary to 8 * 8 partly carries out filtering, then 8 * 4 block boundary is partly carried out filtering, and in vertical direction, block boundary to 8 * 8 partly carries out filtering, then 4 * 8 block boundary is partly carried out filtering.
12. equipment as claimed in claim 12, wherein, described audio-visual codec meets the VC-1 standard of film and television Ssociety of engineers (SMPTE).
13. an audio-visual codec filter apparatus comprises:
Buffer is used to store first filter field;
The row and column buffer is used to store second filter field; With
Filter is used for carrying out according to time sequencing first filter field and second filter field is gone into piece filtering.
14. audio-visual codec filter apparatus as claimed in claim 13, wherein said row and column buffer comprises:
Independent line buffer; With
Independent column buffer.
15. audio-visual codec filter apparatus as claimed in claim 13, wherein, described first filter field comprises macro block.
16. audio-visual codec filter apparatus as claimed in claim 15, wherein, described second filter field comprises the borderline region between the described macro block and second macro block.
17. audio-visual codec filter apparatus as claimed in claim 13, wherein, described audio-visual codec meets the VC-1 standard of film and television Ssociety of engineers (SMPTE).
18. a codec readable memory that is used for according to filtering method control audio-visual codec, described method comprises:
Determine to want filtered a plurality of macro blocks;
Execution is gone into piece filtering to filter field, and described filter field comprises and the borderline region of last filtered macro block at least one and a plurality of macro block in a plurality of macro blocks.
19. codec readable memory as claimed in claim 18, wherein, described audio-visual codec meets the VC-1 standard of film and television Ssociety of engineers (SMPTE).
20. an audio-visual codec filtering method comprises:
For macro block is specified filter field;
Augment described filter field; With
Filter field after augmenting is carried out filtering.
21. method as claimed in claim 20, wherein, augmenting of described filter field comprises: will be included in this filter field with the boundary pixel adjacent pixels of this macro block.
22. method as claimed in claim 21, wherein, augmenting of described filter field comprises: the macro block of previous filtering is included in this filter field.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR46653/04 | 2004-06-22 | ||
| KR20040046653 | 2004-06-22 | ||
| KR55893/04 | 2004-07-19 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1722842A true CN1722842A (en) | 2006-01-18 |
Family
ID=35912736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2005100922132A Pending CN1722842A (en) | 2004-06-22 | 2005-06-22 | The filtering method of audio-visual codec and filter apparatus |
Country Status (2)
| Country | Link |
|---|---|
| KR (1) | KR20050121627A (en) |
| CN (1) | CN1722842A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101193305B (en) * | 2006-11-21 | 2010-05-12 | 安凯(广州)微电子技术有限公司 | Inter-frame prediction data storage and exchange method for video coding and decoding chip |
| CN101115195B (en) * | 2006-07-24 | 2010-08-18 | 同济大学 | Macroblock grade coupled decoding and loop filtering method and apparatus for video code stream |
| CN102939752A (en) * | 2010-04-05 | 2013-02-20 | 三星电子株式会社 | Method and apparatus for encoding video by performing in-loop filtering based on tree-structured data unit, and method and apparatus for decoding video by performing the same |
| CN101742318B (en) * | 2010-01-06 | 2015-01-21 | 威盛电子股份有限公司 | Video decoding and unblocking device and method |
| US9262804B2 (en) | 2010-04-05 | 2016-02-16 | Samsung Electronics Co., Ltd. | Method and apparatus for performing interpolation based on transform and inverse transform |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100753504B1 (en) * | 2005-12-20 | 2007-08-31 | 엘지전자 주식회사 | In-loop Deblocking Filter System and method for filterring data using the same |
| KR100816461B1 (en) * | 2006-04-12 | 2008-03-26 | 주식회사 칩스앤미디어 | Real-time deblocking filter and Method using the same |
| US11284072B2 (en) | 2010-08-17 | 2022-03-22 | M&K Holdings Inc. | Apparatus for decoding an image |
| EP3125561B1 (en) | 2010-08-17 | 2018-03-28 | M&K Holdings Inc. | Method for restoring an intra prediction mode |
| US20130188732A1 (en) * | 2012-01-20 | 2013-07-25 | Qualcomm Incorporated | Multi-Threaded Texture Decoding |
-
2004
- 2004-07-19 KR KR1020040055893A patent/KR20050121627A/en not_active Application Discontinuation
-
2005
- 2005-06-22 CN CNA2005100922132A patent/CN1722842A/en active Pending
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101115195B (en) * | 2006-07-24 | 2010-08-18 | 同济大学 | Macroblock grade coupled decoding and loop filtering method and apparatus for video code stream |
| CN101193305B (en) * | 2006-11-21 | 2010-05-12 | 安凯(广州)微电子技术有限公司 | Inter-frame prediction data storage and exchange method for video coding and decoding chip |
| CN101742318B (en) * | 2010-01-06 | 2015-01-21 | 威盛电子股份有限公司 | Video decoding and unblocking device and method |
| CN102939752A (en) * | 2010-04-05 | 2013-02-20 | 三星电子株式会社 | Method and apparatus for encoding video by performing in-loop filtering based on tree-structured data unit, and method and apparatus for decoding video by performing the same |
| US9262804B2 (en) | 2010-04-05 | 2016-02-16 | Samsung Electronics Co., Ltd. | Method and apparatus for performing interpolation based on transform and inverse transform |
| CN105744273A (en) * | 2010-04-05 | 2016-07-06 | 三星电子株式会社 | Method and apparatus for encoding video, and method and apparatus for decoding video |
| US9390470B2 (en) | 2010-04-05 | 2016-07-12 | Samsung Electronics Co., Ltd. | Method and apparatus for performing interpolation based on transform and inverse transform |
| US9424625B2 (en) | 2010-04-05 | 2016-08-23 | Samsung Electronics Co., Ltd. | Method and apparatus for performing interpolation based on transform and inverse transform |
| US9436975B2 (en) | 2010-04-05 | 2016-09-06 | Samsung Electronics Co., Ltd. | Method and apparatus for performing interpolation based on transform and inverse transform |
| US9547886B2 (en) | 2010-04-05 | 2017-01-17 | Samsung Electronics Co., Ltd. | Method and apparatus for performing interpolation based on transform and inverse transform |
| CN105744273B (en) * | 2010-04-05 | 2018-12-07 | 三星电子株式会社 | The method that video is decoded |
Also Published As
| Publication number | Publication date |
|---|---|
| KR20050121627A (en) | 2005-12-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1722842A (en) | The filtering method of audio-visual codec and filter apparatus | |
| TWI703861B (en) | Method of video decoding at a decoder and apparatus thereof | |
| EP2652954B1 (en) | Video encoding and decoding using transforms | |
| CN1154344C (en) | Method for reducing demand of video frequency decoder processing ability | |
| CN1812576A (en) | Deblocking filters for performing horizontal and vertical filtering of video data simultaneously and methods of operating the same | |
| JP3734492B2 (en) | Image encoding system, image decoding system, and pixel value prediction method for blocks obtained by dividing an image | |
| CN107277539B (en) | Method and apparatus for reducing line cache in context adaptive entropy decoding | |
| JP4680608B2 (en) | Image decoding apparatus and method | |
| CN1725860A (en) | The filtering method that in audio-video codec, uses, equipment and medium | |
| CN1290342C (en) | Apparatus capable of performing both block-matching motion compensation and global motion compensation and method thereof | |
| WO2016196843A1 (en) | Techniques for resource conservation during performance of intra block copy prediction searches | |
| US20050281339A1 (en) | Filtering method of audio-visual codec and filtering apparatus | |
| CN1179574C (en) | Method and device for modifying data in encoded date stream | |
| CN1287598C (en) | Filtering method and apparatus for removing blocking artifacts and/or ringing noise | |
| CN1794814A (en) | Pipelined deblocking filter | |
| CN101031086A (en) | Video-information encoding method and video-information decoding method | |
| CN1717053A (en) | Be used to eliminate filter and method thereof based on the discontinuity of the image of block encoding | |
| CN1767644A (en) | Non-integer pixel sharing for video encoding | |
| CN1708134A (en) | Method and apparatus for estimating motion | |
| EP2807825A1 (en) | Methods and devices for context modeling to enable modular processing | |
| CN1149853C (en) | Coding and decoding device with movement compensation | |
| US9699466B2 (en) | Adaptive reference/non-reference frame determination for video encoding | |
| CN111586405B (en) | Prediction mode rapid selection method based on ALF filtering in multifunctional video coding | |
| CN1489868A (en) | Image information encoding method and encoder, and image information decoding method decoder | |
| US20070147511A1 (en) | Image processing apparatus and image processing 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 |