CN1688164A - Fast motion assessment method based on object edge shape - Google Patents
Fast motion assessment method based on object edge shape Download PDFInfo
- Publication number
- CN1688164A CN1688164A CN 200510026594 CN200510026594A CN1688164A CN 1688164 A CN1688164 A CN 1688164A CN 200510026594 CN200510026594 CN 200510026594 CN 200510026594 A CN200510026594 A CN 200510026594A CN 1688164 A CN1688164 A CN 1688164A
- Authority
- CN
- China
- Prior art keywords
- bab
- piece
- mvps
- shape
- search
- 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.)
- Granted
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
A shape quick motion evaluation method based on an object edge in the video encoding field includes: 1, selecting a shape motion vector prediction value MVP3: taking the mean value of the shape motion vector (MVSI, MVS2, MVS3) adjacent to BAB or the mean value adjacent to the streak block motor vector as the present BAB shape motion vector prediction value (MVP3) 2, the jumping-oven method from the middle: the condition for BAB jumping over the latter search process and finishing the shape motion evaluation values is: if the difference of current BAB and MC-BAB is smaller than or equal to TH, the evaluation process ends, finally MV3=MVP3, 3, selection of the matched error function: balancing BAB matched error by weighted SAD weight, 4, determination of searched zones: searching is carried out in the reference Alpha plane taking the determined position by MVP3 as the center, the zones are set near to VOP border.
Description
Technical field
What the present invention relates to is a kind of method of fast motion assessment, more specifically, relates to a kind of method of the shape fast motion assessment based on target edges, belongs to technical field of video coding.
Background technology
At the development of video coding technique, Motion Picture Experts Group's (MPEG tissue) has formulated the MPEG-4 international standard, and the encoding function to the arbitrary shape object video is provided.In the MPEG-4 standard based on object coding, each object video is described by three category informations usually: shape information, movable information and texture information, wherein shape coding is the key of MPEG-4 based on the object video coding.The core algorithm of shape coding is based on contextual arithmetic coding (CAE), is divided into CAE and interframe CAE in the frame, is respectively applied for the intraframe coding and the interframe encode of shape information.When carrying out the interframe shape coding, in order to improve the code efficiency of CAE, must carry out estimating motion to the shape information of object: find the best matching blocks of present encoding two-value α piece (BAB) in reference figuration information, this is the key of decision CAE code efficiency.Shape estimating motion in the MPEG-4 verification model adopts global search (FS), and the whole candidate's reference blocks by in the match search zone find best matching blocks, has realized accurate movement valuation.But this method will take the amount of calculation of shape coding device more than 90%, becomes the highest module of computation complexity in the encoder.
Find through literature search prior art, people such as Tsung-Han Tsai are in IEEE video technology circuit and system's transactions (IEEE Transactions on Circuits and Systems for Video Technology), Vol.14, NO.6, pp.908-913, in " a kind of quick two-value estimating motion algorithm of MPEG-4 shape coding " delivered on the Jun.2004 (A Fast Binary Motion Estimation Algorithm for MPEG-4Shape Coding), proposition is based on shape fast motion assessment (BPB) method of boundary pixel, improved the speed of shape estimating motion greatly, the core concept of this method is to utilize the boundary pixel of object video, and the improvement of two aspects is arranged:
The firstth, with the absolute difference of weighting and (WSAD) replace the absolute value of difference and (SAD) weigh matching error, can avoid shape motion vector residual error (MVDs) excessive and cause number of coded bits to increase; This method has been carried out weighting simultaneously to SAD and MVDs, with the matching error that its weighted sum WSAD weighs the BAB piece, in order to obtain best weighted effect, need choose suitable weight coefficient W1 and W2.But when weighing the matching error of BAB piece, SAD is still topmost decisive factor, thus the size of MVDs with due regard to only, and determine two weight coefficients than determining weight coefficient difficulties.
Second, this method thinks that the object outline in the current BAB piece should be overlapped with the object outline in the reference BAB piece, therefore in the hunting zone of motion vector, can skip a large amount of search points, these object outline and object outline in the current BAB piece of skipping in the search point corresponding reference BAB piece are not overlapping.But when the border of video object plane (VOP) changed greatly, optimal match point was not often at the boundary pixel place of VOP, and only carry out the BAB piece this moment at the boundary pixel place of VOP coupling is just improper, and this also is the drawback of this method.
Summary of the invention
The present invention is directed to the deficiencies in the prior art, a kind of method of the shape fast motion assessment (OBB) based on target edges has been proposed, make its real-time coding that can realize shape, improved the speed of shape estimating motion greatly, and the code efficiency of shape does not reduce.
The present invention is achieved by the following technical solutions, and concrete steps are as follows:
(1) shape motion vector prediction value MVPs chooses
If it is accurate more that shape motion vector prediction value MVPs chooses, then MVPs is just near more apart from final shape motion vector MVs, and promptly MVDs=MVs-MVPs is more little, so the number of coded bits of MVDs is just few more; And in the estimating motion process of back, can MVPs be the center, search for MVs with littler hunting zone, this just can significantly reduce search point, accelerates the speed of shape estimating motion, and can obviously not increase the number of coded bits of VOP shape.For the BAB piece, because belong to same object, so the shape motion vector MVs between the contiguous BAB piece has very strong correlation.In existing shape estimating motion method, all choose candidate motion vector tabulation { MVs1, MVs2, MVs3, MV1, MV2, first effective motion vector is as the shape motion vector prediction value MVPs of current BAB piece among the MV3}, but this method does not make full use of the strong correlation of shape motion vector between a plurality of contiguous BAB pieces, and has just utilized the shape motion vector information of one of them contiguous BAB piece simply.
Therefore, among the present invention, with the block-shaped motion vector { MVs1 of the contiguous BAB of current BAB piece, MVs2, the intermediate value of MVs3} or contiguous texture block motion vector { MV1, MV2, the intermediate value of MV3} is as the shape motion vector prediction value MVPs of current BAB piece, thereby make MVDs become littler, the number of coded bits of MVDs is also just corresponding to be reduced, and has improved the coding efficiency of VOP shape.
(2) skipping method in the middle of
Determine after the MVPs that the initial motion compensation block MC-BAB of current BAB piece in reference Alpha plane also just determined.In the verification model of MPEG-4, if the matching error of each 4 * 4 sub-piece correspondence is all smaller or equal to 16 * AlphaTH in the BAB piece, illustrate that matching precision has satisfied requirement, then skip the search procedure of back, the shape estimating motion finishes.But the BAB piece that this method is skipped is also few, and the speed of whole shape estimating motion can not get tangible raising.If the BAB piece is skipped, MVDs does not need coding, has saved the number of coded bits of VOP shape.In order to improve the speed of shape estimating motion, under the situation of guaranteeing not obvious increase VOP shape coding bit number, can suitably reduce matching precision, make more BAB piece skip the search procedure of back.
Therefore, the BAB piece is skipped the search procedure of back to the inventive method and the condition that finishes the shape estimating motion changes into: if the matching error between current BAB piece and the MC-BAB piece is smaller or equal to threshold value TH, then shape estimating motion process finishes, final MVs=MVPs.Use this middle skipping method, more BAB piece has been skipped the estimating motion process of back, has saved a large amount of amounts of calculation; And the number of coded bits of VOP shape does not increase, and has descended on the contrary.The improvement of this VOP shape coding bit number, the not enough number of coded bits that causes of matching precision rises in the time of can being used for remedying the back because of fast motion assessment.
(3) the matching error function chooses
In the shape coding process, need coding MVDs, so the big young pathbreaker of MVDs directly influences the number of coded bits of VOP shape.In the shape estimating motion algorithm of MPEG-4 verification model, the absolute value of usefulness motion compensation difference and SAD promptly weigh the matching error of BAB piece as the matching error function with SAD.When obtaining optimum Match by SAD, if MVDs is bigger, can increase the number of coded bits of shape equally, therefore in shape estimating motion process, also should consider the size of MVDs.In order to reduce the influence of bigger MVDs to the VOP shape coding, the inventive method is weighed the matching error of BAB piece by the SADweight of weighting, can between SAD and MVDs, keep a balance, in the minimum SAD of search, guarantee that MVDs can be very not big, thereby the number of coded bits of VOP shape is reduced as far as possible.
(4) region of search determines
After the shape motion vector prediction value MVPs that has determined current BAB piece, the position of determining with MVPs in reference Alpha plane is that the center is searched for, and obtains final shape motion vector MVs.In order to accelerate the speed of shape estimating motion, in the method for the invention, the position of determining with MVPs is the center, search MVs in ± 4 scopes.Because the MVDs overwhelming majority is all in ± 4 scope, so this method can't obviously increase the number of coded bits of VOP shape.For three types of BAB piece, transparent BAB piece does not need coding, the opaque BAB piece in the general all match reference Alpha planes of opaque BAB piece, and its matching error is 0, in interrupt method, belong to the BAB piece of skipping, need not to continue the search procedure of back.Therefore, in the process of search MVs, mainly be the coupling of carrying out boundary B AB piece.
In the methods of the invention, the border characteristic that makes full use of object is carried out the shape estimating motion of boundary B AB piece.For current boundary B AB piece, corresponding reference BAB piece nearly all is a boundary block in reference Alpha plane.In view of this consideration, when carrying out the shape estimating motion of boundary B AB piece, only need carry out the coupling of BAB piece, in the hunting zone, just can significantly reduce the number of coupling BAB piece like this at the VOP boundary vicinity on reference Alpha plane.And for determining of very crucial VOP boundary region of search, the present invention has adopted a kind of method of determining the width of VOP boundary region of search flexibly, has taken into account the speed and the code efficiency of estimating motion.
The inventive method has been utilized the correlation of shape motion vector between the contiguous shape BAB piece more fully, improved the precision of shape motion vector prediction value, the predicted value that is the shape motion vector is nearer apart from the net shape motion vector, therefore can in littler scope, search for final shape motion vector, thereby improved the speed of shape estimating motion greatly, and the number of coded bits of VOP shape increases significantly not; Border characteristic when this method has also made full use of the shape estimating motion---this also is the core concept of this method, the region of search is set in the boundary vicinity of VOP, improved the speed of shape estimating motion so greatly, and the matching precision of boundary B AB piece does not significantly decrease, so the number of coded bits of VOP shape does not have tangible increase yet.Simulation result shows, the present invention is better than existing shape estimating motion method: shape estimating motion speed substantially exceeds the shape estimating motion speed of existing algorithm, its average search count have only that algorithm search is counted in the MPEG-4 verification model 0.82%, and the code efficiency of shape does not reduce.
Description of drawings
Fig. 1 the inventive method flow chart.
Embodiment
Content below in conjunction with accompanying drawing and method provides specific embodiment.
As shown in the figure, the present invention can be divided into following step:
(1) shape motion vector prediction value MVPs's chooses
{ { intermediate value of MV3} is as the shape motion vector prediction value MVPs of current BAB piece for MV1, MV2 for the intermediate value of MVs3} or contiguous texture block motion vector for MVs1, MVs2 with the block-shaped motion vector of contiguous BAB of current BAB piece.Concrete grammar is as follows:
If contiguous BAB piece { there are effective value in MVs1, MVs2, then among the MVs3}
(1) if having only an effective MVs, MVPs equals this MVs;
(2) if having only two effective MVs, then make the 3rd invalid MVs equal (0,0), MVPs=Median (MVs1, MVs2, MVs3); [annotate: intermediate value is got in Median () expression.]
(3) if three MVs are effectively, then MVPs=Median (MVs1, MVs2, MVs3).
If MVs1, MVs2, MVs3} are invalid, but contiguous texture block { there are effective value in MV1, MV2, then among the MV3}
(1) if having only an effective MV, MVPs equals this MV;
(2) if having only two effective MV, then make the 3rd invalid MV equal (0,0), MVPs=Median (MV1, MV2, MV3);
(3) if three MV are effectively, then MVPs=Median (MV1, MV2, MV3);
If MVs1, and MVs2, MVs3, MV1, MV2, MV3} are invalid, then MVPs=(0,0).
After having determined MVPs, calculate the matching error between the definite reference BAB piece of current BAB piece and MVPs, if this matching error is not more than pre-set threshold, illustrate that the reference BAB piece that MVPs determines has reached matching precision, then skip the estimating motion process of back, final MVs=MVPs, MVDs=0.If the BAB piece that MVPs determines can not meet the requirements of matching precision, then need to continue the estimating motion process of back, search for final MVs, and coding MVDs.
Table 1 has provided when adopting the improved median prediction method of the method for choosing MVPs in the MPEG-4 verification model and the present invention respectively, the number of the BAB piece of being skipped.Cycle tests is " Akiyo " of 300 frame CIF forms, " Weather " and " Stefan " sequence.As can be seen from the table, after adopting the improved median prediction method of the present invention, the BAB piece of skipping has increased, and illustrates that the intermediate value of the contiguous BAB piece of use MVs is more accurate as MVPs, thereby makes the matching error of many BAB piece correspondences all be not more than pre-set threshold.Because the BAB piece of skipping does not need to continue the estimating motion process of back, so the BAB piece of skipping is many more, the speed of shape estimating motion is just fast more.Table 1 has provided when adopting different MVPs choosing method simultaneously, and the average number of coded bits of VOP shape is represented with Bits/VOP.As can be seen from the table, use the inventive method, the number of coded bits of VOP shape has reduced, mainly be because: the BAB piece that skip (1) increases, and its corresponding MVDs is 0, does not need the MVDs that encodes; The MVDs of the BAB piece correspondence of (2) not skipping diminishes, and more be distributed near (0,0), so the number of coded bits of MVDs reduces.
The BAB piece of skipping when table 1 adopts MVPs difference choosing method and the comparison of shape coding bit number
| The sequence name | Choose the method for MVPs in the verification model | Improved median prediction method | ||||||
| The BAB that skips | ??％ | ?Bits/VOP | ??％ | The BAB that skips | ??％ | Bits/VOP | ??％ | |
| ?Akiyo | ????42734 | ?100 | ??567.17 | ?100 | ????43967 | ?102.89 | ??544.17 | ?95.97 |
| ?Weather | ????23348 | ?100 | ??609.82 | ?100 | ????24108 | ?103.26 | ??582.29 | ?95.49 |
| ?Stefan | ????1440 | ?100 | ??696.33 | ?100 | ????1492 | ?103.61 | ??682.98 | ?98.08 |
In the cataloged procedure of shape motion vector, only need coding transmission MVDs, rather than coding transmission MVs; In decoding end, after choosing MVPs and decoding MVDs, can calculate MVs=MVPs+MVDs immediately.Because MPEG-4 standardization, the standard decoder of MPEG-4 is as MVPs with first effective MVs of being close to the BAB piece, therefore for standard decoder can be correctly decoded, when transmitting MVDs, should calculate the intermediate value of contiguous BAB piece MVs and the difference between first effective MVs in MVDs to decoder.That is: establish the MVPs=MVmedian that adopts the inventive method to choose, the MVPs=MVfirst that adopts the method in the verification model to choose, the MVDs=MVDs+ that then calculates (MVmedian-MVfirst).This moment, the number of coded bits of MVDs may increase to some extent, but it is that increase and not obvious, and owing to adopt the algorithm that describes below also can save some number of coded bits, so can't obviously increase the shape coding bit number after the employing entire method, for most of cycle tests, number of coded bits also can reduce.
(2) skipping method in the middle of
Determine after the MVPs that the initial motion compensation block MC-BAB of current BAB piece in reference Alpha plane also just determined.The BAB piece is skipped the search procedure of back and the condition that finishes the shape estimating motion is: if the matching error between current BAB piece and the MC-BAB piece smaller or equal to threshold value TH, then shape estimating motion process finishes, final MVs=MVPs.TH=255 * 10 are chosen in test (cycle tests: " Akiyo " of 300 frame CIF forms, " Container ", " Coastguard ", " Weather ", " News ", " Stefan ") by experiment.Find that by emulation for different cycle testss, the ratio that the BAB piece of being skipped accounts for whole BAB pieces is 19.5～63%, this just can obviously accelerate the speed of whole shape estimating motion.Table 2 provided the middle skipping method that adopts the MPEG-4 verification model respectively and the present invention improved in the middle of during skipping method, the number of the BAB piece of being skipped and the average number of coded bits of VOP shape.As can be seen from the table, use the middle skipping method of the present invention, more BAB piece has been skipped the estimating motion process of back, has saved a large amount of amounts of calculation; And the number of coded bits of VOP shape does not increase, and has descended on the contrary.
The BAB piece of skipping when table 2 adopts different middle skipping method and the comparison of shape coding bit number
| The sequence name | The middle skipping method of verification model | Skipping method in the middle of improved | ||||||
| The BAB that skips | ?％ | ?Bits/VOP | ?％ | The BAB that skips | ?％ | ??Bits/VOP | ?％ | |
| ??Akiyo | ??42734 | ?100 | ??567.17 | ?100 | ?50183 | ?117.43 | ??512.25 | ?90.32 |
| ??Weather | ??23348 | ?100 | ??609.82 | ?100 | ?31503 | ?134.93 | ??593.03 | ?97.25 |
| ??Stefan | ??1440 | ?100 | ??696.33 | ?100 | ??3672 | ?255.00 | ??695.20 | ?99.84 |
(3) the matching error function chooses
Weigh the matching error of BAB piece by the SADweight of weighting, between SAD and MVDs, keep a balance, in the minimum SAD of search, guarantee that MVDs can be very not big, thereby the number of coded bits of VOP shape is reduced as far as possible.Adopt following formula to weigh the matching error of BAB piece:
SAD
weight＝SAD+0.5×(|MVDs_x|+|MVDs_y|)
MVDs_x and MVDs_y are respectively level and the vertical component of MVDs in the formula.
Table 3 has provided when adopting based on weighting WSAD method in the boundary pixel algorithm and weighting SADweight method of the present invention respectively, the average number of coded bits of VOP shape.Method of the present invention is with close based on the effect of the weighting WSAD method in the boundary pixel algorithm, and the present invention is much better.Only need a multiplication when the present invention calculates weighting SADweight in addition, and need two multiplication during based on boundary pixel algorithm computation weighting WSAD, this has wasted certain amount of calculation in real-time coding.
The comparison of shape coding bit number when table 3 is chosen different matching error functions
| The sequence name | The WSAD method | ??????SAD weightMethod | ||
| ??Bits/VOP | ??％ | ??Bits/VOP | ?％ | |
| ??Akiyo | ??567.17 | ??100 | ??567.01 | ?99.97 |
| ??Weather | ??609.82 | ??100 | ??608.56 | ?99.79 |
| ??Stefan | ??696.33 | ??100 | ??694.12 | ?99.68 |
(4) region of search determines
Behind the shape motion vector prediction value MVPs that has determined current BAB piece, the position of determining with MVPs in reference Alpha plane is the center, search MVs in ± 4 scopes.
In the present invention, the width of VOP boundary region of search is 8, it determines that method is: in reference Alpha plane, with each pixel is the summit, form one 8 * 8 piece, if this piece is a boundary block, promptly the pixel in the existing VOP has the outer pixel of VOP again in the piece, and then this is in the boundary search zone of VOP; If this piece is transparent block or opaque, then this is not in the boundary search zone of VOP.If expectation further improves the precision of shape estimating motion, can be 16 * 16 to the maximum, but the speed of shape estimating motion will reduce to 8 * 8 amplifications; If expectation further improves the speed of shape estimating motion, can 8 * 8 be dwindled, but the precision of shape estimating motion will reduce, influence code efficiency.
Determine after the MVPs,, need to continue the final MVs of search for the BAB piece of not skipping,
Its search procedure is as follows:
Before carrying out BAB piece estimating motion, determine the region of search of boundary B AB piece at the VOP boundary vicinity on reference Alpha plane;
When carrying out the estimating motion of BAB piece, in reference Alpha plane, the position of determining with MVPs is the center, the final MVs of search in ± 4 scopes;
When in ± 4 scopes, searching for MVs,, continue next coupling of searching for point if the search point is then skipped this search point not in predetermined VOP boundary search zone.
Use the different cycle tests of 300 frames to shape estimating motion algorithm in the MPEG-4 verification model, carry out emulation based on shape fast motion assessment algorithm and these three kinds of shape estimating motion algorithms of the inventive method of boundary pixel below, table 4 has provided the statistics of emulation.
Can draw from table 4: the inventive method is all fine to the sequence effect of various type of sports, the speed of its shape estimating motion substantially exceed in the verification model algorithm and based on the shape estimating motion speed of boundary pixel algorithm, average search count have only that algorithm search is counted in the MPEG-4 verification model 0.82%, and the number of coded bits of VOP shape is not significantly increased, in many sequences, the number of coded bits of VOP shape also reduces to some extent, and promptly the code efficiency of shape does not reduce.
The execution of table 4 different shape estimating motion algorithm is (cycle tests is 300 frames) relatively
| The sequence name | ??Container | ?Coastguard | ????Akiyo | ????News | ????Weather | ????Stefan | |
| Form | ????QCIF | ????QCIF | ????CIF | ????CIF | ????CIF | ????CIF | |
| Shape estimating motion algorithm (VM) in the verification model | ????Bits/VOP | ????128.65 | ????217.89 | ????567.17 | ????569.65 | ????609.82 | ????696.33 |
| ????％ | ????100 | ????100 | ????100 | ????100 | ????100 | ????100 | |
| The search point | ????2101036 | ????3228415 | ????8102104 | ????8910911 | ????10986320 | ????9564043 | |
| ????％ | ????100 | ????100 | ????100 | ????100 | ????100 | ????100 | |
| Shape fast motion assessment algorithm (BPB) based on boundary pixel | ????Bits/VOP | ????128.03 | ????218.93 | ????568.99 | ????570.56 | ????614.82 | ????703.56 |
| ????％ | ????99.52 | ????100.48 | ????100.32 | ????100.16 | ????100.82 | ????101.04 | |
| The search point | ????125424 | ????210375 | ????254264 | ????283598 | ????404421 | ????399361 | |
| ????％ | ????5.97 | ????6.52 | ????3.14 | ????3.18 | ????3.68 | ????4.18 | |
| The present invention is based on the shape fast motion assessment algorithm (OBB) of target edges | ????Bits/VOP | ????122.52 | ????219.21 | ????511.86 | ????539.28 | ????606.85 | ????706.60 |
| ????％ | ????95.24 | ????100.61 | ????90.25 | ????94.67 | ????99.51 | ????101.47 | |
| The search point | ????4949 | ????38666 | ????1516 | ????1274 | ????64778 | ????275060 | |
| ????％ | ????0.24 | ????1.20 | ????0.02 | ????0.01 | ????0.59 | ????2.88 |
In sum, the present invention is better than other method, and its advantage is: (1) does not suppose that the shape of VOP is constant, and is therefore all fine to the effect of various sequences; (2) avoided the deficiency of above-mentioned existing method; (3) utilized the correlation of shape motion vector between the contiguous shape piece more fully, the intermediate value of the shape motion vector of the contiguous BAB piece of utilization is as the shape motion vector prediction value of current BAB piece, improved the precision of predicted value, the accuracy of predicted value has guaranteed can dwindle the hunting zone greatly under the situation of not obvious increase VOP shape coding bit number; Border characteristic when (4) having made full use of the shape estimating motion is set in the boundary vicinity of VOP to the region of search, has improved the speed of shape estimating motion greatly, makes between VOP shape coding bit number and the shape estimating motion speed to have kept balance; (5) matching precision of boundary shape piece can't significantly decrease, and it is close that VOP shape coding bit number keeps, and the shape coding bit number also can reduce to some extent in many video sequences, and promptly the code efficiency of shape does not reduce.This method can improve the shape estimating motion speed of existing algorithm greatly on the basis that keeps shape coding efficient.
Claims (5)
1, a kind of method of the shape fast motion assessment based on target edges is characterized in that step is as follows:
(1) shape motion vector prediction value MVPs chooses
{ { intermediate value of MV3} is as the shape motion vector prediction value MVPs of current BAB piece for MV1, MV2 for the intermediate value of MVs3} or contiguous texture block motion vector for MVs1, MVs2 with the block-shaped motion vector of contiguous BAB of current BAB piece;
(2) skipping method in the middle of
Determine after the MVPs, the initial motion compensation block MC-BAB of current BAB piece in reference Alpha plane also just determined, the BAB piece skips the search procedure of back and the condition that finishes the shape estimating motion is: if the matching error between current BAB piece and the MC-BAB piece is smaller or equal to threshold value TH, then shape estimating motion process finishes, final MVs=MVPs;
(3) the matching error function chooses
Weigh the matching error of BAB piece by the SADweight of weighting, between SAD and MVDs, keep a balance;
(4) region of search determines
After the shape motion vector prediction value MVPs that has determined current BAB piece, the position of determining with MVPs in reference Alpha plane is the center, searches in ± 4 scopes, and the region of search is set at the VOP boundary, obtains final shape motion vector MVs.
2, the method for the shape fast motion assessment based on target edges according to claim 1 is characterized in that, described step (1), and it is as follows that shape motion vector prediction value MVPs chooses concrete grammar:
If contiguous BAB piece { there are effective value in MVs1, MVs2, then among the MVs3}
(1) if having only an effective MVs, MVPs equals this MVs,
(2) if having only two effective MVs, then make the 3rd invalid MVs equal (0,0), MVPs=Median (MVs1, MVs2, MVs3),
(3) if three MVs are effectively, then MVPs=Median (MVs1, MVs2, MVs3);
If MVs1, MVs2, MVs3} are invalid, but contiguous texture block { there are effective value in MV1, MV2, then among the MV3}
(1) if having only an effective MV, MVPs equals this MV,
(2) if having only two effective MV, then make the 3rd invalid MV equal (0,0), MVPs=Median (MV1, MV2, MV3),
(3) if three MV are effectively, then MVPs=Median (MV1, MV2, MV3);
If MVs1, and MVs2, MVs3, MV1, MV2, MV3} are invalid, then MVPs=(0,0);
After having determined MVPs, calculate the matching error between the definite reference BAB piece of current BAB piece and MVPs, if this matching error is less than or equal to pre-set threshold, illustrate that the reference BAB piece that MVPs determines has reached matching precision, then skip the estimating motion process of back, final MVs=MVPs, MVDs=0; If the BAB piece that MVPs determines can not meet the requirements of matching precision, then need to continue the estimating motion process of back, search for final MVs, and coding MVDs.
3, the method for the shape fast motion assessment based on target edges according to claim 1 is characterized in that, in the described step (3), adopts following formula to weigh the matching error of BAB piece:
SAD
weight＝SAD+0.5×(|MVDs_x|+|MVDs_y|)
MVDs_x and MVDs_y are respectively level and the vertical component of MVDs in the formula.
4, the method for the shape fast motion assessment based on target edges according to claim 1, it is characterized in that, in the described step (4), the width of VOP boundary region of search is 8, and definite method of region of search is: in reference Alpha plane, be the summit with each pixel, form one 8 * 8 piece, if this piece is a boundary block, promptly the pixel in the existing VOP has the outer pixel of VOP again in the piece, and then this is in the boundary search zone of VOP; If this piece is transparent block or opaque, then this by amplifying the precision that further improve the shape estimating motions to 8 * 8, is 16 * 16 to the maximum outside the boundary search zone of VOP; By the speed of dwindling further raising shape estimating motion to 8 * 8.
5, the method for the shape fast motion assessment based on target edges according to claim 1 is characterized in that, determines after the MVPs, for the BAB piece of not skipping, needs to continue the final MVs of search, and its search procedure is as follows:
Before carrying out BAB piece estimating motion, determine the region of search of boundary B AB piece at the VOP boundary vicinity on reference Alpha plane;
When carrying out the estimating motion of BAB piece, in reference Alpha plane, the position of determining with MVPs is the center, the final MVs of search in ± 4 scopes;
When in ± 4 scopes, searching for MVs,, continue next coupling of searching for point if the search point is then skipped this search point outside predetermined VOP boundary search zone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100265944A CN100337482C (en) | 2005-06-09 | 2005-06-09 | Fast motion assessment method based on object edge shape |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100265944A CN100337482C (en) | 2005-06-09 | 2005-06-09 | Fast motion assessment method based on object edge shape |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1688164A true CN1688164A (en) | 2005-10-26 |
| CN100337482C CN100337482C (en) | 2007-09-12 |
Family
ID=35306204
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2005100265944A Expired - Fee Related CN100337482C (en) | 2005-06-09 | 2005-06-09 | Fast motion assessment method based on object edge shape |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100337482C (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101867812A (en) * | 2010-04-16 | 2010-10-20 | 中山大学 | Method for estimating and predicting video data compression motion by using edge effect to predict video data compression motion |
| CN101631241B (en) * | 2008-07-14 | 2011-05-25 | 晨星软件研发(深圳)有限公司 | Image processing method and device |
| CN101505419B (en) * | 2008-02-05 | 2011-06-15 | 晨星半导体股份有限公司 | Image processing apparatus for determining moving vector and related method thereof |
| CN101374235B (en) * | 2007-08-24 | 2012-05-02 | 大唐移动通信设备有限公司 | Method and apparatus for estimating rapid block motion of video encoding |
| WO2012071949A1 (en) * | 2010-11-29 | 2012-06-07 | Mediatek Inc. | Method and apparatus for derivation of mv/mvp candidate for inter/skip/merge modes |
| WO2012174990A1 (en) * | 2011-06-24 | 2012-12-27 | Mediatek Inc. | Method and apparatus for removing redundancy in motion vector predictors |
| CN101494780B (en) * | 2008-01-25 | 2013-01-23 | 联发科技股份有限公司 | Method and device for video processing |
| CN103002196A (en) * | 2011-09-09 | 2013-03-27 | 联咏科技股份有限公司 | Method for estimating prediction motion vector |
| CN110213588A (en) * | 2019-06-25 | 2019-09-06 | 浙江大华技术股份有限公司 | Airspace Candidate Motion information acquisition method, device, codec and storage device |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH1169356A (en) * | 1997-08-25 | 1999-03-09 | Mitsubishi Electric Corp | Dynamic image encoding system and dynamic image decoding system |
| KR100523052B1 (en) * | 2002-08-30 | 2005-10-24 | 한국전자통신연구원 | Object base transmission-receive system and method, and object-based multiview video encoding apparatus and method for supporting the multi-display mode |
| CN100382600C (en) * | 2004-04-22 | 2008-04-16 | 上海交通大学 | Detection method of moving object under dynamic scene |
-
2005
- 2005-06-09 CN CNB2005100265944A patent/CN100337482C/en not_active Expired - Fee Related
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101374235B (en) * | 2007-08-24 | 2012-05-02 | 大唐移动通信设备有限公司 | Method and apparatus for estimating rapid block motion of video encoding |
| CN101494780B (en) * | 2008-01-25 | 2013-01-23 | 联发科技股份有限公司 | Method and device for video processing |
| CN101505419B (en) * | 2008-02-05 | 2011-06-15 | 晨星半导体股份有限公司 | Image processing apparatus for determining moving vector and related method thereof |
| CN101631241B (en) * | 2008-07-14 | 2011-05-25 | 晨星软件研发(深圳)有限公司 | Image processing method and device |
| CN101867812A (en) * | 2010-04-16 | 2010-10-20 | 中山大学 | Method for estimating and predicting video data compression motion by using edge effect to predict video data compression motion |
| WO2012071949A1 (en) * | 2010-11-29 | 2012-06-07 | Mediatek Inc. | Method and apparatus for derivation of mv/mvp candidate for inter/skip/merge modes |
| WO2012174990A1 (en) * | 2011-06-24 | 2012-12-27 | Mediatek Inc. | Method and apparatus for removing redundancy in motion vector predictors |
| US10063876B2 (en) | 2011-06-24 | 2018-08-28 | Hfi Innovation Inc. | Method and apparatus for removing redundancy in motion vector predictors |
| US10097850B2 (en) | 2011-06-24 | 2018-10-09 | Hfi Innovation Inc. | Method and apparatus for removing redundancy in motion vector predictors |
| CN103002196A (en) * | 2011-09-09 | 2013-03-27 | 联咏科技股份有限公司 | Method for estimating prediction motion vector |
| CN110213588A (en) * | 2019-06-25 | 2019-09-06 | 浙江大华技术股份有限公司 | Airspace Candidate Motion information acquisition method, device, codec and storage device |
| CN110213588B (en) * | 2019-06-25 | 2021-07-13 | 浙江大华技术股份有限公司 | Spatial domain candidate motion information acquisition method and device, coder-decoder and storage device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100337482C (en) | 2007-09-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100337482C (en) | Fast motion assessment method based on object edge shape | |
| CN1291604C (en) | Calculating method for predicting motion vector | |
| CN100348051C (en) | An enhanced in-frame predictive mode coding method | |
| CN1225127C (en) | A coding/decoding end bothway prediction method for video coding | |
| CN1186938C (en) | Decoder with digital image stability function and image stability method | |
| US8467448B2 (en) | Apparatus and method for fast intra/inter macro-block mode decision for video encoding | |
| CN100527848C (en) | Motion vector detecting device and motion vector detecting method | |
| CN1791224A (en) | Self-adaptive block searching range rapid motion estimating method based on H.264 | |
| CN1675848A (en) | Method and apparatus for performing high quality fast predictive motion search | |
| CN101420617B (en) | Motion estimation searching method for cross hexagon | |
| CN101394566B (en) | Cross rhombic motion estimation searching method | |
| CN1189037C (en) | Motion estimation | |
| CN1275469C (en) | Method for pridicting sortable complex in frame | |
| CN1956547A (en) | Motion vector estimating device and motion vector estimating method | |
| CN1708132A (en) | Method and apparatus for predicting motion | |
| CN1777289A (en) | Method for speeding up motion estimation utilizing selective prediction | |
| CN1656515A (en) | Unit for and method of estimating a current motion vector | |
| CN1604650A (en) | Method for hierarchical motion estimation | |
| CN100340116C (en) | Motion estimating method with graded complexity | |
| CN1325220A (en) | Motion-vector coding method | |
| CN1656514A (en) | Unit for and method of estimating a current motion vector | |
| CN1758759A (en) | Switching detection method of video scene | |
| CN1181691C (en) | Vidio motion estimation method | |
| CN103096064B (en) | The method and relevant device of coding and reconstructed pixel block | |
| CN1913632A (en) | Motion picture encoding method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20070912 Termination date: 20130609 |