CN103702127B - Motion estimation search range Forecasting Methodology based on motion vector dependency and system - Google Patents
Motion estimation search range Forecasting Methodology based on motion vector dependency and system Download PDFInfo
- Publication number
- CN103702127B CN103702127B CN201310744084.5A CN201310744084A CN103702127B CN 103702127 B CN103702127 B CN 103702127B CN 201310744084 A CN201310744084 A CN 201310744084A CN 103702127 B CN103702127 B CN 103702127B
- Authority
- CN
- China
- Prior art keywords
- motion vector
- mvph
- mvpv
- srbasic
- icount
- 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.)
- Active
Links
Abstract
The invention provides motion estimation search range Forecasting Methodology based on motion vector dependency and system, the method comprising the steps of: S1, presets maximum search scope SRmaxWith minimum search range SRmin;The fitting coefficient of predicting unit PU under S2, calculating different mode;S3, pattern according to current PU, obtain corresponding fitting coefficient;S4, pattern according to current PU, obtain corresponding motion vector predictor and available spatially and temporally correlation candidate motion vector;S5, calculation of motion vectors predict the cumulative difference of sub and each available spatially and temporally correlation candidate motion vector;S6, according to the fitting coefficient in S2, the motion vector predictor in S4, the cumulative difference in S5, calculate basic search scope;S7, according to basic search scope, maximum search scope SR set in advancemaxWith minimum search range SRminDetermine final hunting zone SRk.This invention can dynamically adjust hunting zone, thus reaches reduce amount of calculation and shorten the purpose of search time.
Description
Technical field
The present invention relates to technical field of video processing, particularly relate to based on motion vector dependency
Motion estimation search range Forecasting Methodology and system.
Background technology
Along with the development of technology, fine definition, that the video of high frame per second becomes people gradually is new
Demand.High efficiency Video coding (HEVC, High Efficiency Video Coding) is by moving
State motion picture expert group version (MPEG, Moving Pictures Experts Group) and international telecommunication connection
Alliance's telecommunication standardsization tissue (ITU-T, ITU-T for ITU Telecommunication
Standardization Sector) Video Coding Experts group (VCEG, Video Coding Experts
Group) Video coding collectively constituted combines group (JCT-VC, Joint Collaborative Team
On Video Coding) video compression standard of new generation developed jointly.HEVC is meant to ensure that and regards
Frequently can be by Bit-Rate Reduction to previous generation video encoding standard on the premise of image quality does not declines
H.264/MPEG-4AVC 50%.But, the cost that bit rate declines is then coding complexity
Being increased sharply of degree.Can be used to reduce huge amount of calculation despite many methods, but
Also there is a need to the work of many to further speed up the speed of coding.HEVC have employed code tree
Unit (CTU, Coding Tree Unit) structure.Can comprise inside each CTU one or
Multiple coding units (CU, Coding Unit), and each CU correspond to predicting unit (PU,
Prediction Unit) and converter unit (TU, Transform Unit).Estimation (ME,
Motion Estimation) time, current PU is by needing search in encoded reference frame
The part mated most.
At present, the mode of PU Optimizing Search process is mainly in a fixing hunting zone
Preferable position is obtained, the problem that this method exists by search:
1, the motion conditions of current PU itself is not accounted for, also cannot be according to current PU's
Motion conditions reasonably limits the scope of search;
2, the concordance of the motion of current PU and CU around is not accounted for, also cannot be certainly
The adjustment hunting zone adapted to;
Summary of the invention
For the deficiencies in the prior art, the present invention provides intraframe coding based on pattern pretreatment excellent
Change method and system so that can stably accelerate coding rate in HEVC encoding-operation process.
For achieving the above object, the present invention is achieved by the following technical programs:
Motion estimation search range Forecasting Methodology based on motion vector dependency, including following step
Rapid:
S1, preset maximum search scope SRmaxWith minimum search range SRmin;
The fitting coefficient ah of predicting unit PU under S2, calculating different modek,avkAnd bhk,bvk,
Wherein the pattern sum of predicting unit PU is Q, 0≤k < Q;
S3, pattern k according to current PU, obtain corresponding fitting coefficient ahk,avkAnd bhk,bvk;
S4, pattern k according to current PU, obtain corresponding motion vector predictor
(MVPhk,MVPvk) and available spatially and temporally correlation candidate motion vector
(MVChki,MVCvki), wherein, iCountkRepresent available spatially and temporally correlation candidate motion
Total number of vector, 0≤i < iCountk;
S5, calculation of motion vectors prediction son (MVPhk,MVPvk) and each available spatially and temporally phase
Close candidate motion vector (MVChki,MVCvki) cumulative difference (Δ MVThk,ΔMVTvk);
S6, according to the fitting coefficient ah in S2k,avkAnd bhk,bvk, the motion vector in S4 is pre-
Survey son (MVPhk,MVPvk), cumulative difference (the Δ MVTh in S5k,ΔMVTvk), calculate basic search scope
SRBasick;
S7, according to basic search scope SRBasic in S6kWith the maximum search scope in S1
SRmaxWith minimum search range SRminDetermine final hunting zone SRk。
Further, described step S2 includes:
S21, determine all patterns of predicting unit PU;
PU in S22, acquisition cycle testsjkThe available candidate spatially and temporally moving relevant fortune
Moving vector (MVChijk,MVCvijk), 0≤k < Q, j represents jth PU under kth kind pattern,
Q is the assemble mode number of predicting unit;
S23, acquisition PUjkMotion vector predictor (MVPhjk,MVPvjk) and actual motion vector
(MVhjk,MVvjk);
S24, calculation of motion vectors prediction son (MVPhjk,MVPvjk) and each available spatial domain and time
Territory correlation candidate motion vector (MVChijk,MVCvijk) difference, after taking absolute value sue for peace:
Wherein, iCountjkRepresent available spatially and temporally correlation candidate motion vector
Number;
S25, calculating PUjkActual motion vector (MVhjk,MVvjk) and motion vector predictor
(MVPhjk,MVPvjk) difference, and take absolute value:
ΔMVhjk=| MVhjk-MVPhjk|
ΔMVvjk=| MVvjk-MVPvjk|
S26, structure following two equation group:
ahk×|MVPhjk|+bhk×ΔMVThjk=Δ MVhjk
avk×|MVPvjk|+bvk×ΔMVTvjk=Δ MVvjk
Least square fitting algorithm is utilized to be calculated ahk,avkAnd bhk,bvk。
Further, described step S5 includes:
Calculation of motion vectors prediction son (MVPhk,MVPvk) and each available being spatially and temporally correlated with
Candidate motion vector (MVChki,MVCvki) difference, after taking absolute value sue for peace:
Wherein, iCountkRepresent the number of available spatially and temporally correlation candidate motion vector.
Further, described step S6 includes:
According to the fitting coefficient ah in S2k,avkAnd bhk,bvk, motion vector predictor in S4
(MVPhk,MVPvk), cumulative difference (the Δ MVTh in S5k,ΔMVTvk), calculating basic search scope:
SRhk-basic=| ahk×ΔMVThk+bhk×|MVPhk||
SRvk-basic=| avk×ΔMVTvk+bvk×|MVPvk||。
SRBasick=SRhk-basic+SRvk-basic
Further, described step S7 includes:
According to basic search scope SRBasickWith maximum search scope SR set in advancemax?
Little hunting zone SRminDetermine final hunting zone SRkFor:
Motion estimation search range prognoses system based on motion vector dependency, this system includes:
Searched in advance range set module, is used for presetting maximum search scope SRmaxSearch with minimum
Rope scope SRmin;
Fitting coefficient computing module, for calculating the fitting coefficient of predicting unit PU under different mode
ahk,avkAnd bhk,bvk, wherein the pattern sum of predicting unit PU is Q, 0≤k < Q;
The fitting coefficient acquisition module of associative mode, for pattern k according to current PU, obtains
Corresponding fitting coefficient ahk,avkAnd bhk,bvk;
Prediction of associative mode and associated motion vector acquisition module, for according to current PU's
Pattern k, obtains corresponding motion vector predictor (MVPhk,MVPvk) and available spatial domain and time
Territory correlation candidate motion vector (MVChki,MVCvki), wherein, iCountkRepresent available spatial domain and
Total number of time domain correlation candidate motion vector, 0≤i < iCountk;
Correlation calculations module, for calculation of motion vectors prediction son (MVPhk,MVPvk) and each can
Spatially and temporally correlation candidate motion vector (MVChki,MVCvki) cumulative difference
(ΔMVThk,ΔMVTvk);
Basic search range computation module, for the fitting coefficient acquisition module according to associative mode
In fitting coefficient ahk,avkAnd bhk,bvk, prediction of associative mode and associated motion vector obtain
Motion vector predictor (MVPh in delivery blockk,MVPvk), the cumulative difference in correlation calculations module
(ΔMVThk,ΔMVTvk), calculate basic search scope SRBasick;
Final hunting zone determines module, for according to the base in basic search range computation module
This hunting zone SRBasickWith maximum search scope SR in searched in advance range set modulemax
With minimum search range SRminDetermine final hunting zone SRk。
Further, described fitting coefficient computing module includes:
PU pattern determining unit, for determining all patterns of predicting unit PU;
Associated motion vector acquiring unit, is used for obtaining PU in cycle testsjkAvailable spatial domain and
Candidate motion vector (the MVCh that temporal motion is relevantijk,MVCvijk), < Q, j represent kth to 0≤k
Jth PU under the pattern of kind, k is the pattern of current prediction unit;
Prediction and actual motion vector acquiring unit, be used for obtaining PUjkMotion vector predictor
(MVPhjk,MVPvjk) and actual motion vector (MVhjk,MVvjk);
Correlation calculations unit, for calculation of motion vectors prediction son (MVPhjk,MVPvjk) and each
Individual available spatially and temporally correlation candidate motion vector (MVChijk,MVCvijk) difference, take absolute value
Rear summation:
Wherein, iCountjkRepresent the number of available spatially and temporally correlation candidate motion vector;
Difference computational unit, is used for calculating PUjkActual motion vector (MVhjk,MVvjk) and motion
Vector predictor (MVPhjk,MVPvjk) difference, and take absolute value:
Coefficient Fitting unit, is used for constructing following two equation group:
ahk×|MVPhjk|+bhk×ΔMVThjk=Δ MVhjk
avk×|MVPvjk|+bvk×ΔMVTvjk=Δ MVvjk
Least square fitting algorithm is utilized to be calculated ahk,avkAnd bhk,bvk。
Wherein, described correlation calculations module is for calculation of motion vectors prediction son (MVPhk,MVPvk)
With each available spatially and temporally correlation candidate motion vector (MVChki,MVCvki) difference, take absolutely
Sue for peace to after value:
Wherein, iCountkRepresent the number of available spatially and temporally correlation candidate motion vector.
Further, described basic search range computation module is for the matching according to associative mode
Fitting coefficient ah in coefficient acquisition modulek,avkAnd bhk,bvk, prediction of associative mode and phase
Close the motion vector predictor (MVPh in motion vector acquisition modulek,MVPvk), correlation calculations mould
Cumulative difference (Δ MVTh in blockk,ΔMVTvk), calculate basic search scope SRBasick:
SRhk-basic=| ahk×ΔMVThk+bhk×|MVPhk||
SRvk-basic=| avk×ΔMVTvk+bvk×|MVPvk||。
SRBasick=SRhk-basic+SRvk-basic
Further, described final hunting zone determines that module is for according to basic search scope
SRBasickWith maximum search scope SR set in advancemaxWith minimum search range SRminDetermine
Whole hunting zone SRk:
The present invention has a following beneficial effect:
The present invention is by utilizing the motion vector predictor (MVPh of current PUk,MVPvk), available
Spatially and temporally correlation candidate motion vector (MVChki,MVCvki), fitting coefficient ahk,avkWith
bhk,bvkDynamically adjust basic search scope SRBasick, finally by maximum search model set in advance
Enclose SRmax, minimum search range SRminWith basic search scope SRBasic dynamically adjustedkCommon true
Fixed final hunting zone SRk.Correlation candidate motion is spatially and temporally gone up owing to present invention utilizes
The motion vector predictor of the current PU of vector sum combines the motion estimation search predicting current PU
Scope, therefore, it is possible to adjust hunting zone real-time dynamicly according to practical situation, thus reduces meter
Calculation amount, decreases search time simultaneously.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below
The accompanying drawing used required in embodiment or description of the prior art will be briefly described, aobvious and
Easily insight, the accompanying drawing in describing below is some embodiments of the present invention, common for this area
From the point of view of technical staff, on the premise of not paying creative work, it is also possible to according to these accompanying drawings
Obtain other accompanying drawing.
Fig. 1 is motion estimation search model based on motion vector dependency in the embodiment of the present invention 1
Enclose the flow chart of Forecasting Methodology;
Fig. 2 is motion estimation search model based on motion vector dependency in the embodiment of the present invention 2
Enclose the flow chart of Forecasting Methodology;
Fig. 3 is the relative of the candidate motion vector that the spatial domain described in the embodiment of the present invention 2 is relevant
Position view;
Fig. 4 is the relative of the candidate motion vector that the time domain described in the embodiment of the present invention 2 is relevant
Position view;
Fig. 5 is motion estimation search model based on motion vector dependency in the embodiment of the present invention 3
Enclose the structural representation of prognoses system.
Detailed description of the invention
For making the purpose of the embodiment of the present invention, technical scheme and advantage clearer, below will knot
Close the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear,
Complete description, it is clear that described embodiment be a part of embodiment of the present invention rather than
Whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not having
Make the every other embodiment obtained under creative work premise, broadly fall into present invention protection
Scope.
Embodiment 1
The embodiment of the present invention 1 proposes motion estimation search range based on motion vector dependency
Forecasting Methodology, sees Fig. 1, comprises the steps:
Step 101: preset maximum search scope SRmaxWith minimum search range SRmin。
Step 102: calculate the fitting coefficient ah of predicting unit PU under different modek,avkWith
bhk,bvk, wherein the pattern sum of predicting unit PU is Q, 0≤k < Q.
Step 103: according to pattern k of current PU, obtain corresponding fitting coefficient ahk,avkWith
bhk,bvk。
Step 104: according to pattern k of current PU, obtain corresponding motion vector predictor
(MVPhk,MVPvk) and available spatially and temporally correlation candidate motion vector (MVChki,MVCvki),
Wherein, iCountkRepresent total number of available spatially and temporally correlation candidate motion vector,
0≤i<iCountk。
Step 105: calculation of motion vectors prediction son (MVPhk,MVPvk) and each available spatial domain and
Time domain correlation candidate motion vector (MVChki,MVCvki) cumulative difference (Δ MVThk,ΔMVTvk)。
Step 106: according to the fitting coefficient ah in step 102k,avkAnd bhk,bvk, step 104
In motion vector predictor (MVPhk,MVPvk), the cumulative difference in step 105
(ΔMVThk,ΔMVTvk), calculate basic search scope SRBasick。
Step 107: according to basic search scope SRBasickWith maximum search scope set in advance
SRmaxWith minimum search range SRminDetermine final hunting zone SRk。
Visible, the embodiment of the present invention is by utilizing the motion vector predictor of current PU
(MVPhk,MVPvk), available spatially and temporally correlation candidate motion vector (MVChki,MVCvki), intend
Syzygy number ahk,avkAnd bhk,bvkDynamically adjust basic search scope SRBasick, finally by setting in advance
Fixed maximum search scope SRmax, minimum search range SRminWith the basic search scope dynamically adjusted
SRBasickJointly determine final hunting zone SRk.To sum up describe visible, owing to the present invention is real
Execute example to make use of and spatially and temporally go up correlation candidate motion vector and estimate to the motion predicting current PU
Meter hunting zone, therefore, it is possible to adjust hunting zone real-time dynamicly according to practical situation, thus
Reach to reduce the purpose of amount of calculation.
Embodiment 2
A preferred embodiment of the more detailed description present invention is carried out below by embodiment 2
Realize process.Seeing Fig. 2, this process comprises the steps:
Step 201: preset maximum search scope SRmaxWith minimum search range SRmin。
In this step, SRmaxValue be referred to the fixing search scope of prior art, example
As, the fixing search scope before not using the present invention to be optimized is set to 64, then
SR hereinmaxIt is also configured as 64;Due to the hunting zone that likely obtains when being predicted very
Little close to zero, if directly using the hunting zone of prediction, the result of search may be bad,
Now use and limit a minimum hunting zone SRmin, can be by the SR limitedminIt is suitable to come
Increase hunting zone and obtain a more stable result.General SRminTake 4 be one proper
Value.
Step 202: determine all patterns of predicting unit PU.
In this step, the mould of all different predicting unit (PU, Prediction Unit) is determined
Formula: with reference to HM software, such as, for the CTU of 64x64, pattern Q=20 of PU, divides
It is not:
4×8,4×16,4×8,8×8,8×16,
8×32,12×16,16×4,16×8,16×12,
16×16,16×32,24×32,32×8,32×16,
32×24,32×32,32×64,64×32,64×64。
Due to the motion conditions difference in the region that the PU of different mode is comprised, so for difference
The PU of pattern uses different fitting coefficients to be obtained in that reasonable effect.
Step 203: obtain PU in cycle testsjkAvailable relevant time of spatially and temporally moving
Select motion vector (MVChijk,MVCvijk), 0≤k < Q, j represents the jth under kth kind pattern
PU, Q are the pattern sum of predicting unit.
In this step, suitable cycle tests is determined, it is thus achieved that different mode in cycle tests
PUjkAvailable candidate motion vector (the MVCh spatially and temporally moving relevantijk,MVCvijk), including
Current PUjkThe left side, spatial domain (L), lower-left (LD), top (U), upper right (RU) and a left side
The motion vector (as shown in Figure 3) of upper (LU) 5 CU and time domain bottom right (RD) or
The motion vector (as shown in Figure 4) of 1 CU in both centers (C), wherein time domain fortune
In moving vector, the priority of bottom right is higher than center, and records relevant time of spatially and temporally moving
Select the number iCount of motion vectorjk, wherein j represents jth PU, and 0≤k < Q represents
The PU of k kind pattern.In implementation process, owing to PU location is different, actual available
Number iCount of candidate motion vectorjkAlso different.Available candidate motion vector depends on after obtaining
Secondary leaving in stores before position, can record the individual of final available candidate motion vector simultaneously
Number iCountjk;And disabled candidate motion vector is set to (0,0), leave storage position in
Below.When calculating owing to simply using above iCountjkIndividual candidate motion vector, does not use
It is stored in candidate motion vector below, obtains several motion vectors, the party so actual
Method can be correct process.
Step 204: obtain PUjkMotion vector predictor (MVPhjk,MVPvjk) and actual motion to
Amount (MVhjk,MVvjk)。
In this step, HEVC test model HM can centered by motion vector predictor,
SRmaxIn the range of search obtain optimal matched position, this optimal position is exactly actual motion
Vector.This step is the process of estimation in fact: before estimation only know motion to
Amount prediction son, by motion vector predictor obtain reality motion vector, record this two
Class value.
Step 205: calculation of motion vectors prediction son (MVPhjk,MVPvjk) and each available spatial domain
With time domain correlation candidate motion vector (MVChijk,MVCvijk) cumulative difference.
In this step, solve in the following way:
Step 206: calculate PUjkActual motion vector (MVhjk,MVvjk) and motion vector predictor
(MVPhjk,MVPvjk) difference, and take absolute value.
In this step, solve as follows:
ΔMVhjk=| MVhjk-MVPhjk|
ΔMVvjk=| MVvjk-MVPvjk|
Step 207: equationof structure group, utilizes least square fitting algorithm to obtain fitting coefficient
ahk,avkAnd bhk,bvk。
In this step, structure following two equation group:
ahk×|MVPhjk|+bhk×ΔMVThjk=Δ MVhjk
avk×|MVPvjk|+bvk×ΔMVTvjk=Δ MVvjk
Wherein k represents different PU patterns, and j represents all values belonging to pattern k, utilizes
A young waiter in a wineshop or an inn takes advantage of fitting algorithm to be calculated ahk,avkAnd bhk,bvk。
Step 208: according to pattern k of current PU, obtain corresponding motion vector predictor
(MVPhk,MVPvk) and available spatially and temporally correlation candidate motion vector (MVChki,MVCvki),
Wherein, iCountkRepresent total number of available spatially and temporally correlation candidate motion vector,
0≤i<iCountk。
In this step, it is thus achieved that the motion vector predictor (MVPh of current PUk,MVPvk) and can
Spatially and temporally correlation candidate motion vector (MVChki,MVCvki), including current PU spatial domain
The left side (L), lower-left (LD), top (U), upper right (RU) and upper left (LU) 5
Both the motion vector (as shown in Figure 3) of CU and time domain bottom right (RD) or center (C)
In the motion vector (as shown in Figure 4) of 1 CU, wherein bottom right in temporal motion vector
Priority higher than center, wherein i represent all available spatially and temporally correlation candidate motions to
In amount one, and record the number iCount of motion vectork;
Step 209: calculation of motion vectors prediction son (MVPhk,MVPvk) and each available spatial domain and
Time domain correlation candidate motion vector (MVChki,MVCvki) cumulative difference (Δ MVThk,ΔMVTvk)。
In this step, cumulative poor (Δ MVTh is solved in such a wayk,ΔMVTvk):
Step 210: according to the fitting coefficient ah in step 207k,avkAnd bhk,bvk, step 208
In motion vector predictor (MVPhk,MVPvk), the cumulative difference in step 209
(ΔMVThk,ΔMVTvk), calculate basic search scope SRBasick。
In this step, basic search scope SRBasic is solved in such a wayk:
SRhk-basic=| ahk×ΔMVThk+bhk×|MVPhk||
SRvk-basic=| avk×ΔMVTvk+bvk×|MVPvk||
SRBasick=SRhk-basic+SRvk-basic
(ΔMVThk,ΔMVTvk) illustrate the motion phase of current PU and the most relevant CU
Guan Xing: if two values are smaller, illustrates that current PU motion conditions is to the most relevant
The ratio of CU more consistent, hunting zone should be less;Otherwise, they motion conditions are described not
Unanimously, hunting zone should become big.(MVPhk,MVPvk) illustrate motion feelings possible for current PU
Condition: two values are smaller, illustrates that the motion ratio of current PU itself is shallower or background is without motion,
The reduction that so hunting zone can be suitable;Otherwise, illustrate that the motion ratio of current PU itself is more acute
Strong, then hunting zone should be bigger.
Step 211: according to basic search scope SRBasickWith maximum search scope set in advance
SRmaxWith minimum search range SRminDetermine final hunting zone SRk。
In this step, carry out in such a way solving final hunting zone SRk:
Size by real-time adjustment SR, it is possible to the effective time reducing integral point search.
Test result shows, the method can reduce by the integral point search time of 20% to 55%, simultaneously
Ensure that the distortion rate of signal is at below 0.08dB.
The embodiment of the present invention is by utilizing the motion vector predictor (MVPh of current PUk,MVPvk)、
Available spatially and temporally correlation candidate motion vector (MVChki,MVCvki), fitting coefficient ahk,avk
And bhk,bvkDynamically adjust basic search scope SRBasick, finally by maximum search set in advance
Scope SRmax, minimum search range SRminWith basic search scope SRBasic dynamically adjustedkJointly
Determine final hunting zone SRk.To sum up describe visible, owing to the embodiment of the present invention make use of sky
On territory and time domain, correlation candidate motion vector predicts the motion estimation search range of current PU, because of
This can adjust hunting zone real-time dynamicly according to practical situation, thus has reached to reduce calculating
Amount and the purpose shortening search time.
Embodiment 3
The embodiment of the present invention 3 also proposed motion estimation search model based on motion vector dependency
Enclosing prognoses system, see Fig. 5, this system includes:
Hunting zone presets module 501, is used for presetting maximum search scope SRmax?
Little hunting zone SRmin;
Fitting coefficient computing module 502, for calculating the matching of predicting unit PU under different mode
Coefficient ahk,avkAnd bhk,bvk, wherein the pattern sum of predicting unit PU is Q, 0≤k < Q;
The fitting coefficient acquisition module 503 of associative mode, for pattern k according to current PU,
Obtain corresponding fitting coefficient ahk,avkAnd bhk,bvk;
Prediction of associative mode and associated motion vector acquisition module 504, for according to current
Pattern k of PU, obtains corresponding motion vector predictor (MVPhk,MVPvk) and available sky
Territory and time domain correlation candidate motion vector (MVChki,MVCvki), wherein, iCountkExpression can use
Spatially and temporally total number of correlation candidate motion vector, 0≤i < iCountk;
Correlation calculations module 505, for calculation of motion vectors prediction son (MVPhk,MVPvk) and every
Individual available spatially and temporally correlation candidate motion vector (MVChki,MVCvki) cumulative difference
(ΔMVThk,ΔMVTvk);
Basic search range computation module 506, obtains mould for the fitting coefficient according to associative mode
Fitting coefficient ah in blockk,avkAnd bhk,bvk, prediction of associative mode and associated motion vector
Motion vector predictor (MVPh in acquisition modulek,MVPvk), adding up in correlation calculations module
Difference (Δ MVThk,ΔMVTvk), calculate basic search scope SRBasick;
Final hunting zone determines module 507, for according in basic search range computation module
Basic search scope SRBasickWith the maximum search scope in searched in advance range set module
SRmaxWith minimum search range SRminDetermine final hunting zone SRk。
Wherein, described fitting coefficient computing module 502 includes:
PU pattern determining unit 5020, for determining all patterns of predicting unit PU:
In this unit, with reference to HM software, such as, for the CTU of 64x64, PU's
Pattern Q=20, is respectively as follows:
4×8,4×16,4×8,8×8,8×16,
8×32,12×16,16×4,16×8,16×12,
16×16,16×32,24×32,32×8,32×16,
32×24,32×32,32×64,64×32,64×64。
Due to the motion conditions difference in the region that the PU of different mode is comprised, so for difference
The PU of pattern uses different fitting coefficients to be obtained in that reasonable effect.
Associated motion vector acquiring unit 5021, is used for obtaining PU in cycle testsjkAvailable
Candidate motion vector (the MVCh spatially and temporally moving relevantijk,MVCvijk), 0≤k < Q, j
Representing jth PU under kth kind pattern, Q is the assemble mode number of predicting unit;
In this unit, it is first determined suitably cycle tests, then in acquisition cycle tests not
PU with patternjkThe available relevant candidate motion vector that spatially and temporally moves
(MVChijk,MVCvijk), including current PUjkThe left side, spatial domain (L), lower-left (LD), top (U),
The motion vector (as shown in Figure 3) of upper right (RU) and upper left (LU) 5 CU and
The motion vector of 1 CU in both time domain bottom right (RD) or center (C) is (such as Fig. 4
Shown in), wherein in temporal motion vector the priority of bottom right higher than center, and record spatial domain and
The number iCount of the candidate motion vector that temporal motion is relevantjk, wherein j represents jth PU,
0≤k < Q represents the PU of kth kind pattern.In implementation process, due to PU location
Difference, number iCount of actual available candidate motion vectorjkAlso different.Available candidate's fortune
Moving vector leaves in before storage position after obtaining successively, can record final available time simultaneously
Select number iCount of motion vectorjk;And disabled candidate motion vector is set to (0,0),
Leave in after storage position.When calculating owing to simply using above iCountjkIndividual candidate transports
Moving vector, does not has to use the candidate motion vector being stored in below, so actual acquisition is several
Individual motion vector, the process that the method can be correct.
Prediction and actual motion vector acquiring unit 5022, be used for obtaining PUjkMotion vector
Prediction son (MVPhjk,MVPvjk) and actual motion vector (MVhjk,MVvjk);
Correlation calculations unit 5023, for calculation of motion vectors prediction son (MVPhjk,MVPvjk) and
Each available spatially and temporally correlation candidate motion vector (MVChijk,MVCvijk) difference, take absolutely
Sue for peace to after value:
Wherein, iCountjkRepresent the number of available spatially and temporally correlation candidate motion vector;
Difference computational unit 5024, is used for calculating PUjkActual motion vector (MVhjk,MVvjk) and
Motion vector predictor (MVPhjk,MVPvjk) difference, and take absolute value:
Coefficient Fitting unit 5025, is used for constructing following two equation group:
ahk×|MVPhjk|+bhk×ΔMVThjk=Δ MVhjk
avk×|MVPvjk|+bvk×ΔMVTvjk=Δ MVvjk
Least square fitting algorithm is utilized to be calculated ahk,avkAnd bhk,bvk。
Wherein, described correlation calculations module 505 is for calculation of motion vectors prediction
(MVPhk,MVPvk) and each available spatially and temporally correlation candidate motion vector
(MVChki,MVCvki) difference, after taking absolute value sue for peace:
Wherein, iCountkRepresent the number of available spatially and temporally correlation candidate motion vector.
Wherein, described basic search range computation module 506 is for the matching system according to associative mode
Fitting coefficient ah in number acquisition modulek,avkAnd bhk,bvk, prediction of associative mode is with relevant
Motion vector predictor (MVPh in motion vector acquisition modulek,MVPvk), correlation calculations module
In cumulative difference (Δ MVThk,ΔMVTvk), calculate basic search scope SRBasick:
SRhk-basic=| ahk×ΔMVThk+bhk×|MVPhk||
SRvk-basic=| avk×ΔMVTvk+bvk×|MVPvk||。
SRBasick=SRhk-basic+SRvk-basic
Wherein, described final hunting zone determines that module 507 is for according to basic search scope
SRBasickWith maximum search scope SR set in advancemaxWith minimum search range SRminDetermine
Whole hunting zone SRk:
System described in the embodiment of the present invention, utilizes the motion vector predictor of current PU
(MVPhk,MVPvk), available spatially and temporally correlation candidate motion vector (MVChki,MVCvki), intend
Syzygy number ahk,avkAnd bhk,bvkDynamically adjust basic search scope SRBasick, finally by setting in advance
Fixed maximum search scope SRmax, minimum search range SRminWith the basic search scope dynamically adjusted
SRBasickJointly determine final hunting zone SRk.To sum up describe visible, owing to make use of sky
On territory and time domain, correlation candidate motion vector predicts the motion estimation search range of current PU, because of
This can adjust hunting zone real-time dynamicly according to practical situation, thus reduces amount of calculation, with
Time also imply that and decrease search time.
Above example is merely to illustrate technical scheme, is not intended to limit;Although ginseng
Being described in detail the present invention according to previous embodiment, those of ordinary skill in the art should
Work as understanding: the technical scheme described in foregoing embodiments still can be modified by it, or
Person carries out equivalent to wherein portion of techniques feature;And these amendments or replacement, do not make phase
The essence answering technical scheme departs from the spirit and scope of various embodiments of the present invention technical scheme.
Claims (10)
1. motion estimation search range Forecasting Methodology based on motion vector dependency, its feature exists
In, comprise the following steps:
S1, preset maximum search scope SRmaxWith minimum search range SRmin;
The fitting coefficient ah of predicting unit PU under S2, calculating different modek,avkAnd bhk,bvk,
Wherein the pattern sum of predicting unit PU is Q, 0≤k < Q;
S3, pattern k according to current PU, obtain corresponding fitting coefficient ahk,avkAnd bhk,bvk;
S4, pattern k according to current PU, obtain corresponding motion vector predictor
(MVPhk,MVPvk) and available spatially and temporally correlation candidate motion vector
(MVChki,MVCvki), wherein, iCountkRepresent available spatially and temporally correlation candidate motion
Total number of vector, 0≤i < iCountk;
S5, calculation of motion vectors prediction son (MVPhk,MVPvk) and each available spatially and temporally phase
Close candidate motion vector (MVChki,MVCvki) cumulative difference (Δ MVThk,ΔMVTvk);
S6, according to the fitting coefficient ah in S2k,avkAnd bhk,bvk, the motion vector in S4 is pre-
Survey son (MVPhk,MVPvk), cumulative difference (the Δ MVTh in S5k,ΔMVTvk), calculate basic search scope
SRBasick;
S7, according to basic search scope SRBasic in S6kWith the maximum search scope in S1
SRmaxWith minimum search range SRminDetermine final hunting zone SRk。
Method the most according to claim 1, it is characterised in that described step S2 includes:
S21, determine all patterns of predicting unit PU;
PU in S22, acquisition cycle testsjkThe available candidate spatially and temporally moving relevant fortune
Moving vector (MVChijk,MVCvijk), 0≤k < Q, j represents jth PU under kth kind pattern,
Q is the assemble mode number of predicting unit;
S23, acquisition PUjkMotion vector predictor (MVPhjk,MVPvjk) and actual motion vector
(MVhjk,MVvjk);
S24, calculation of motion vectors prediction son (MVPhjk,MVPvjk) and each available spatial domain and time
Territory correlation candidate motion vector (MVChijk,MVCvijk) difference, after taking absolute value sue for peace:
Wherein, iCountjkRepresent available spatially and temporally correlation candidate motion vector
Number;
S25, calculating PUjkActual motion vector (MVhjk,MVvjk) and motion vector predictor
(MVPhjk,MVPvjk) difference, and take absolute value:
ΔMVhjk=| MVhjk-MVPhjk|
ΔMVvjk=| MVvjk-MVPvjk|
S26, structure following two equation group:
ahk×|MVPhjk|+bhk×ΔMVThjk=Δ MVhjk
avk×|MVPvjk|+bvk×ΔMVTvjk=Δ MVvjk
Least square fitting algorithm is utilized to be calculated ahk,avkAnd bhk,bvk。
Method the most according to claim 1, it is characterised in that described step S5 includes:
Calculation of motion vectors prediction son (MVPhk,MVPvk) and each available being spatially and temporally correlated with
Candidate motion vector (MVChki,MVCvki) difference, after taking absolute value sue for peace:
Wherein, iCountkRepresent the number of available spatially and temporally correlation candidate motion vector.
Method the most according to claim 1, it is characterised in that described step S6 includes:
According to the fitting coefficient ah in S2k,avkAnd bhk,bvk, motion vector predictor in S4
(MVPhk,MVPvk), cumulative difference (the Δ MVTh in S5k,ΔMVTvk), calculating basic search scope:
SRhk-basic=| ahk×ΔMVThk+bhk×|MVPhk||
SRvk-basic=| avk×ΔMVTvk+bvk×|MVPvk||。
SRBasick=SRhk-basic+SRvk-basic
Method the most according to claim 1, it is characterised in that described step S7 includes:
According to basic search scope SRBasickWith maximum search scope SR set in advancemax?
Little hunting zone SRminDetermine final hunting zone SRkFor:
6. motion estimation search range prognoses system based on motion vector dependency, its feature exists
In, this system includes:
Searched in advance range set module, is used for presetting maximum search scope SRmaxSearch with minimum
Rope scope SRmin;
Fitting coefficient computing module, for calculating the fitting coefficient of predicting unit PU under different mode
ahk,avkAnd bhk,bvk, wherein the pattern sum of predicting unit PU is Q, 0≤k < Q;
The fitting coefficient acquisition module of associative mode, for pattern k according to current PU, obtains
Corresponding fitting coefficient ahk,avkAnd bhk,bvk;
Prediction of associative mode and associated motion vector acquisition module, for according to current PU's
Pattern k, obtains corresponding motion vector predictor (MVPhk,MVPvk) and available spatial domain and time
Territory correlation candidate motion vector (MVChki,MVCvki), wherein, iCountkRepresent available spatial domain and
Total number of time domain correlation candidate motion vector, 0≤i < iCountk;
Correlation calculations module, for calculation of motion vectors prediction son (MVPhk,MVPvk) and each can
Spatially and temporally correlation candidate motion vector (MVChki,MVCvki) cumulative difference
(ΔMVThk,ΔMVTvk);
Basic search range computation module, for the fitting coefficient acquisition module according to associative mode
In fitting coefficient ahk,avkAnd bhk,bvk, prediction of associative mode and associated motion vector obtain
Motion vector predictor (MVPh in delivery blockk,MVPvk), the cumulative difference in correlation calculations module
(ΔMVThk,ΔMVTvk), calculate basic search scope SRBasick;
Final hunting zone determines module, for according to the base in basic search range computation module
This hunting zone SRBasickWith maximum search scope SR in searched in advance range set modulemax
With minimum search range SRminDetermine final hunting zone SRk。
System the most according to claim 6, it is characterised in that described fitting coefficient calculates
Module includes:
PU pattern determining unit, for determining all patterns of predicting unit PU;
Associated motion vector acquiring unit, is used for obtaining PU in cycle testsjkAvailable spatial domain and
Candidate motion vector (the MVCh that temporal motion is relevantijk,MVCvijk), < Q, j represent kth to 0≤k
Jth PU under the pattern of kind, k is the pattern of current prediction unit;
Prediction and actual motion vector acquiring unit, be used for obtaining PUjkMotion vector predictor
(MVPhjk,MVPvjk) and actual motion vector (MVhjk,MVvjk);
Correlation calculations unit, for calculation of motion vectors prediction son (MVPhjk,MVPvjk) and each
Individual available spatially and temporally correlation candidate motion vector (MVChijk,MVCvijk) difference, take absolute value
Rear summation:
Wherein, iCountjkRepresent the number of available spatially and temporally correlation candidate motion vector;
Difference computational unit, is used for calculating PUjkActual motion vector (MVhjk,MVvjk) and motion
Vector predictor (MVPhjk,MVPvjk) difference, and take absolute value:
ΔMVhjk=| MVhjk-MVPhjk|;
ΔMVvjk=| MVvjk-MVPvjk|
Coefficient Fitting unit, is used for constructing following two equation group:
ahk×|MVPhjk|+bhk×ΔMVThjk=Δ MVhjk
avk×|MVPvjk|+bvk×ΔMVTvjk=Δ MVvjk
Least square fitting algorithm is utilized to be calculated ahk,avkAnd bhk,bvk。
System the most according to claim 6, it is characterised in that described correlation calculations mould
Block is for calculation of motion vectors prediction son (MVPhk,MVPvk) and each available spatially and temporally phase
Close candidate motion vector (MVChki,MVCvki) difference, after taking absolute value sue for peace:
Wherein, iCountkRepresent the number of available spatially and temporally correlation candidate motion vector.
System the most according to claim 6, it is characterised in that described basic search scope
Computing module is for according to the fitting coefficient ah in the fitting coefficient acquisition module of associative modek,avk
And bhk,bvk, prediction and the motion vector in associated motion vector acquisition module of associative mode
Prediction son (MVPhk,MVPvk), cumulative difference (the Δ MVTh in correlation calculations modulek,ΔMVTvk), meter
Calculate basic search scope SRBasick:
SRhk-basic=| ahk×ΔMVThk+bhk×|MVPhk||
SRvk-basic=| avk×ΔMVTvk+bvk×|MVPvk||。
SRBasick=SRhk-basic+SRvk-basic
System the most according to claim 6, it is characterised in that described final search model
Enclose and determine that module is for according to basic search scope SRBasickWith maximum search model set in advance
Enclose SRmaxWith minimum search range SRminDetermine final hunting zone SRk:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310744084.5A CN103702127B (en) | 2013-12-30 | 2013-12-30 | Motion estimation search range Forecasting Methodology based on motion vector dependency and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310744084.5A CN103702127B (en) | 2013-12-30 | 2013-12-30 | Motion estimation search range Forecasting Methodology based on motion vector dependency and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103702127A CN103702127A (en) | 2014-04-02 |
CN103702127B true CN103702127B (en) | 2016-10-12 |
Family
ID=50363512
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310744084.5A Active CN103702127B (en) | 2013-12-30 | 2013-12-30 | Motion estimation search range Forecasting Methodology based on motion vector dependency and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103702127B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107087171B (en) * | 2017-05-26 | 2020-02-07 | 中国科学技术大学 | HEVC integer pixel motion estimation method and device |
CN111385583B (en) * | 2018-12-28 | 2022-04-22 | 展讯通信(上海)有限公司 | Image motion estimation method, device and computer readable storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101001377A (en) * | 2007-01-09 | 2007-07-18 | 浙江大学 | Method of H.264 quick motion estimation based on multi-reference frame |
CN101102493A (en) * | 2007-08-07 | 2008-01-09 | 华亚微电子(上海)有限公司 | A method for chip storage applicable motion estimation |
CN103024392A (en) * | 2012-12-26 | 2013-04-03 | 清华大学 | Method and device for intra-frame mode prediction based on two-dimensional Hadamard transformation |
EP2597874A1 (en) * | 2010-07-20 | 2013-05-29 | Ntt Docomo, Inc. | Image prediction encoding device, image prediction encoding method, image prediction encoding program, image prediction decoding device, image prediction decoding method, and image prediction decoding program |
-
2013
- 2013-12-30 CN CN201310744084.5A patent/CN103702127B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101001377A (en) * | 2007-01-09 | 2007-07-18 | 浙江大学 | Method of H.264 quick motion estimation based on multi-reference frame |
CN101102493A (en) * | 2007-08-07 | 2008-01-09 | 华亚微电子(上海)有限公司 | A method for chip storage applicable motion estimation |
EP2597874A1 (en) * | 2010-07-20 | 2013-05-29 | Ntt Docomo, Inc. | Image prediction encoding device, image prediction encoding method, image prediction encoding program, image prediction decoding device, image prediction decoding method, and image prediction decoding program |
CN103024392A (en) * | 2012-12-26 | 2013-04-03 | 清华大学 | Method and device for intra-frame mode prediction based on two-dimensional Hadamard transformation |
Non-Patent Citations (2)
Title |
---|
Motion Feature and Hadamard Coefficient-Based Fast Multiple Reference Frame Motion Estimation for H.264;Zhenyu Liu等;《IEEE Transactions on Circuits and Systems for Video Technology》;20080531;第18卷(第5期);620-632 * |
一种快速HEVC帧内预测算法;成益龙等;《电视技术》;20121102;第36卷(第21期);4-7 * |
Also Published As
Publication number | Publication date |
---|---|
CN103702127A (en) | 2014-04-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101534436B (en) | Allocation method of video image macro-block-level self-adaptive code-rates | |
KR101553850B1 (en) | / Video encoding/decoding apparatus and method and apparatus of adaptive overlapped block motion compensation using adaptive weights | |
CN105933717B (en) | A kind of mode decision method and apparatus of coding unit | |
CN103248895B (en) | A kind of quick mode method of estimation for HEVC intraframe coding | |
CN108781284A (en) | The method and device of coding and decoding video with affine motion compensation | |
CN101243691B (en) | Method and apparatus for estimating motion based on block | |
CN104935938A (en) | Inter-frame prediction method in hybrid video coding standard | |
CN102334337A (en) | Apparatus and method for coding/decoding image selectivly using descrete cosine/sine transtorm | |
CN102932642B (en) | Interframe coding quick mode selection method | |
CN107396102B (en) | A kind of inter-frame mode fast selecting method and device based on Merge technological movement vector | |
CN103533359B (en) | One is bit rate control method H.264 | |
CN110312132A (en) | A kind of decoding method, device and its equipment | |
CN107040782B (en) | Video coding overall situation Rate-distortion optimization method based on Lagrangian method | |
CN106454349B (en) | A kind of estimation block matching method based on H.265 Video coding | |
CN108134939A (en) | A kind of method for estimating and device | |
CN104811729B (en) | A kind of video multi-reference frame coding method | |
CN111314698A (en) | Image coding processing method and device | |
CN107071421B (en) | A kind of method for video coding of combination video stabilization | |
CN109672894A (en) | A kind of inter-frame prediction method, device and storage medium | |
CN108989799A (en) | A kind of selection method, device and the electronic equipment of coding unit reference frame | |
CN104702959B (en) | A kind of intra-frame prediction method and system of Video coding | |
CN110177282A (en) | A kind of inter-frame prediction method based on SRCNN | |
CN109996080A (en) | Prediction technique, device and the codec of image | |
CN103702127B (en) | Motion estimation search range Forecasting Methodology based on motion vector dependency and system | |
CN107302701A (en) | Decode interpolation filter type |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220726 Address after: 100084 Tsinghua University, Beijing, Haidian District Patentee after: TSINGHUA University Patentee after: Huawei Cloud Computing Technology Co.,Ltd. Address before: 100084 mailbox, 100084-82 Tsinghua Yuan, Beijing, Haidian District, Beijing Patentee before: TSINGHUA University Patentee before: HUAWEI TECHNOLOGIES Co.,Ltd. |
|
TR01 | Transfer of patent right |