CN102665061A - Motion vector processing-based frame rate up-conversion method and device - Google Patents

Abstract

The embodiment of the invention discloses a motion vector processing-based frame rate up-conversion method and a motion vector processing-based frame rate up-conversion device. The method comprises the following steps of: performing scene detection on an input original video sequence to judge whether a scene is changed or not currently; if a scene change is detected, processing the original video sequence in an interpolation way to obtain processed data; if the scene change is not detected, processing the original video sequence in an extrapolation way to obtain processed data; calculating the processed data by using symmetrical motion estimation to obtain an initial motion vector field; performing motion vector processing on a motion vector of the initial motion vector field to obtain a new motion vector field; and performing adaptive motion compensation processing on a motion vector of the new motion vector field to obtain a frame to be interpolated. By the method and the device according to the embodiment of the invention, calculation complexity and motion estimation complexity can be lowered, the quality of an interpolated frame can be improved, and the requirements of real-time application can be met.

Description

A kind of based on conversion method and device thereof on the frame per second of motion vector processing

Technical field

The present invention relates to technical field of video processing, relate in particular to a kind of based on conversion method and device thereof on the frame per second of motion vector processing.

Background technology

The large-scale application of digital video and multimedia messages has been brought the diversity of display format.Therefore, the conversion between these forms just seems extremely important.Change on the frame per second, (frame rate up-conversion, FRUC), a kind of exactly ability will be hanged down the technology that the frame-rate video sequence is converted into high frame-rate video sequence also to claim the frame per second lifting.Frame per second promotes has wide application, seems more continuously and nature like the motion that can make video content.Frame per second promotes, and mainly is to be used for realizing two conversions between the different frame per second videoscanning forms.The application scenario of FRUC be HDTV (High Definition Television, HDTV).The resolution of HDTV is the highest can to reach 1920*1080, and frame per second can reach 60 frames/s.And present China and the employed PAL of eurovision (Phase Alternating Line) system frame per second are merely 25 frames/s, and the NTSC (national television system committee) that states such as North America use is made as 30 frames/s.Therefore, be HDTV with having system conversion now, must introduce on the frame per second and change the mechanism.In addition, FRUC also can be applied in the low bit rate video coding.Its thought is suitably to reduce the frame per second of original video at coding side, introduces FRUC mechanism in decoding end then and lets frame per second rise to original frame per second size.

Motion compensated interpolation method (MCFI) is that a kind of effect is better among the FRUC, uses algorithm more widely.This algorithm utilizes the correlation between the consecutive frame, and the motion vector of supposing intermediate frame is the half the of front and back two frame motion vectors, thus the interior intermediate frame of inserting.

But when video sequence strenuous exercise perhaps existed scene to switch, the correlation of consecutive frame descended, and this interior method of inserting will lose efficacy.Simultaneously because MCFI is a kind of block-based algorithm, so when a certain block at object edge and when comprising several motion object, the motion vector of this piece is just inaccurate, thus the soft edge of inserting out in causing.And can not find the pixel that is complementary with it at occlusion area and exposed region interior pixel point, the motion vector that obtains of estimation is just incorrect so, such zone with the MCFI method will in the slotting piece that makes mistake, also just caused blocking artifact.Adopt MCFI can strengthen the complexity of computation complexity and estimation, and reduced the quality of high interpolation frame, be not suitable for real-time application.

Summary of the invention

The objective of the invention is to overcome the deficiency of prior art; The invention provides a kind of based on conversion method and device thereof on the frame per second of motion vector processing; Can reduce the complexity of computation complexity and estimation, can improve the quality of interpolation frame, satisfy the requirement of using in real time.

In order to address the above problem, the present invention proposes conversion method on a kind of frame per second of handling based on motion vector, said method comprises:

Original video sequence to input carries out scene detection, detects the current scene change generation that whether has;

If when detecting the scene change generation, then adopt the data after interpolation method is handled and obtained to handle said original video sequence;

If when not detecting the scene change generation, then adopt the data after the extrapolation mode is handled and obtained to handle said original video sequence;

Adopt symmetric motion to estimate the data after handling are calculated, obtain the initial motion vectors field;

Motion vector to said initial motion vectors field carries out the new motion vector field of motion vector processing acquisition;

Motion vector to said new motion vector field carries out adaptive motion compensated processing, obtains to treat interleave.

Preferably; Said original video sequence to input carries out scene detection; Detecting the current step that whether has scene change to take place comprises: adopt frame difference method to detect scene change, according to statistics absolute difference and the number judgment that SAD surpasses the piece of specific threshold whether the scene change generation is arranged.

Preferably; Said employing symmetric motion is estimated the data after handling are calculated; The step that obtains the initial motion vectors field comprises: with the coordinate of treating inserted block in the interpolation frame is benchmark, and in the hunting zone in two reference frames the position of the mobile corresponding match block of symmetry, select minimum that of difference to piece as best matching blocks; To obtain motion vector, form the initial motion vectors field.

Preferably, said motion vector to said initial motion vectors field carries out the step that motion vector handle to obtain new motion vector field and comprises: the corresponding blocks that motion vector MV is labeled as L1 is divided into sub-piece; With adjacent be labeled as L2 the drink piece is merged into bulk, to obtain variable-block; Said variable-block is carried out estimation, obtain new motion vector field.

Preferably; Said motion vector to said new motion vector field carries out adaptive motion compensated processing; Acquisition treats that the step of interleave comprises: adopt forward frame or consequent frame that present frame is carried out interpolation; Perhaps use mode that bidirectional frame fills up that the motion vector of said new motion vector field is carried out adaptive motion compensated processing, obtain to treat interleave.

Correspondingly, it is a kind of based on conversion equipment on the frame per second of motion vector processing that the embodiment of the invention also provides, and said device comprises:

Scene detection module is used for the original video sequence of input is carried out scene detection, detects the current scene change generation that whether has;

The video sequence processing module is used for then adopting the data after interpolation method is handled and obtained to handle said original video sequence if said scene detection module detects scene change when taking place; If when said scene detection module does not detect the scene change generation, then adopt the data after the extrapolation mode is handled and obtained to handle said original video sequence;

The motion vector acquisition module is used to adopt symmetric motion to estimate the data after the said video sequence processing module processing are calculated, and obtains the initial motion vectors field;

Motion vector processing module, the motion vector of the initial motion vectors field that is used for said motion vector acquisition module is obtained are carried out motion vector and are handled and obtain new motion vector field;

Adaptive interpolation filters, the motion vector of the new motion vector field that is used for said motion vector processing module is obtained carries out adaptive motion compensated processing, obtains to treat interleave.

Whether preferably, said scene detection module also is used to adopt frame difference method to detect scene change, have scene change to take place according to adding up absolute difference and the SAD number judgment above the piece of specific threshold.

Preferably; Said motion vector acquisition module also is used for treating that with interpolation frame the coordinate of inserted block is a benchmark; The position of the mobile corresponding match block of symmetry in the hunting zone in two reference frames; Select minimum that of difference to piece as best matching blocks, to obtain motion vector, form the initial motion vectors field.

Preferably; The corresponding blocks that said motion vector processing module also is used for motion vector MV is labeled as L1 is divided into sub-piece, with adjacent be labeled as L2 the drink piece is merged into bulk, to obtain variable-block; Said variable-block is carried out estimation, obtain new motion vector field.

Preferably; Said adaptive interpolation filters also is used to adopt forward frame or consequent frame that present frame is carried out interpolation; Perhaps use mode that bidirectional frame fills up that the motion vector of said new motion vector field is carried out adaptive motion compensated processing, obtain to treat interleave.

In embodiments of the present invention, taked diverse ways to promote the subjective effect of interpolation image respectively 3 of frame rate conversion different stages.At first, utilize scene detection to judge and be to use the method for interior method of inserting or extrapolation that original video sequence is handled, verification is upgraded in the initial motion vectors field that then estimation is obtained, to obtain motion vector (MV) the most reliably.At last select the weights of interpolation filter further to eliminate local blooming adaptively.Scene detection adopts frame difference method, and is simple and practical in frame rate conversion is used.Because it only needs the number of statistics SAD above the size of a certain threshold value, can reduce the complexity of computation complexity and estimation, little to the influence of overall complexity, in addition, can not only improve the quality of interpolation frame, also satisfy the requirement of using in real time simultaneously.

Description of drawings

In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art; To do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below; Obviously, the accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills; Under the prerequisite of not paying creative work, can also obtain other accompanying drawing according to these accompanying drawings.

Fig. 1 is the schematic flow sheet based on conversion method on the frame per second of motion vector processing of the embodiment of the invention;

Fig. 2 is the scene detection sketch map of the embodiment of the invention;

Fig. 3 is that sketch map is estimated in the symmetric motion of the embodiment of the invention;

Fig. 4 is that the structure based on conversion equipment on the frame per second of motion vector processing of the embodiment of the invention is formed sketch map.

Embodiment

To combine the accompanying drawing in the embodiment of the invention below, the technical scheme in the embodiment of the invention is carried out clear, intactly description, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills are not making the every other embodiment that is obtained under the creative work prerequisite, all belong to the scope of the present invention's protection.

In embodiments of the present invention, utilize the scene detection function that the original video sequence of input is detected, detect the current scene change generation that whether has.If detect current scene generation conversion, adopt the mode of extrapolation that original video sequence is handled; If do not detect scene change, the mode of inserting in adopting is handled original video sequence.Estimate that through symmetric motion (bi-directional motion estimation) obtains an initial motion vectors field (MVF).This initial MVF can roughly reflect the movement locus of object; But fringe region, occlusion area and exposed region at object; Some motion vector (MV) might be unreliable, insecure MV carried out obtain reliable MV after the processing such as verification corrigendum.At last, select for use adaptive interpolation filters to handle to obtain treating the output of interleave.

Fig. 1 is the schematic flow sheet based on conversion method on the frame per second of motion vector processing of the embodiment of the invention, and as shown in Figure 1, this method comprises:

S101 carries out scene detection to the original video sequence of importing, and detects the current scene change generation that whether has; If when detecting scene change and taking place, then carry out S102, if, then carry out S103 not if when detecting scene change and taking place;

S102 adopts the data after interpolation method is handled original video sequence and obtained to handle;

S103 adopts the data after the extrapolation mode is handled original video sequence and obtained to handle;

S104 adopts symmetric motion to estimate the data after handling are calculated, and obtains the initial motion vectors field;

S105 carries out motion vector to the motion vector of initial motion vectors field and handles and obtain new motion vector field;

S106 carries out adaptive motion compensated processing to the motion vector of new motion vector field, obtains to treat interleave.Wherein, S101 comprises:

Adopt frame difference method to detect scene change, whether have scene change to take place according to adding up absolute difference and SAD number judgment above the piece of specific threshold.

Below in conjunction with Fig. 2 the scene detection of the embodiment of the invention is elaborated.

Generally speaking, there is very big correlation between the consecutive frame of video sequence.As shown in Figure 2, f _N-3Frame and f _N-1Between the frame, f _N+1Frame and f _N+3All there is very big correlation between the frame.But when the situation of camera lens switching occurring, can there be great difference in two frames that are in switching point usually, such as f _N-1Frame and f _N+1Frame.If the method for inserting in still adopting this moment is used f _N-1Frame and f _N+1Frame is to f _nFrame is predicted, then can't obtain effective interleave.If use extrapolation, promptly use f in this case _N-3Frame and f _N-1Frame or f _N+1Frame and f _N+3The frame frame of extrapolating often can obtain better effect.

In the practical implementation, adopt frame difference method to detect the conversion of scene, through the statistics absolute difference with (Sum of Absolute Difference, the number that SAD) surpasses the piece of a certain given threshold value (specific threshold) has judged whether that scene change produces.For per two consecutive frames of list entries, calculate the sad value of corresponding macro block by formula (1).

$\mathrm{SAD}\left[B_{\mathrm{ij}}\right]=\underset{s\∈B_{\mathrm{ij}}}{\mathrm{\Σ}}\left|f_{n-1}\right[S]-f_{n+1}[S\left]\right|---\left(1\right)$

Wherein, f _N-1, f _N+1Be two consecutive frames, B _IjIt is a macro block in the frame.

Statistics SAD is greater than the macroblock number (SAD_num) of certain threshold value.Suppose f _N-3With f _N-1Between SAD be SAD_num0 greater than the macroblock number of threshold value, f _N-1With f _N+1Between be SAD_num1, f _N+1With f _N+3Between be SAD_num2.

If SAD_num1＜n explains that so the correlation of consecutive frame is very big, so the method for inserting in adopting.If SAD_num1＞n judges the switching that has scene between these two consecutive frames so.Adopt this moment the method for extrapolation to intermediate frame f _nEstimate.

During extrapolation, if satisfy

$\left\{\begin{array}{c}{\mathrm{SAD}\_\mathrm{num}}_0<n\\ \mathrm{SAD}\_{\mathrm{num}}_1>n\\ \mathrm{SAD}\_{\mathrm{num}}_2>n\end{array}\right.---\left(2\right)$

So just utilize front two frame f _N-3, f _N-1Extrapolation present frame f _n

If satisfy following formula

$\left\{\begin{array}{c}{\mathrm{SAD}\_\mathrm{num}}_0>n\\ \mathrm{SAD}\_{\mathrm{num}}_1>n\\ \mathrm{SAD}\_{\mathrm{num}}_2<n\end{array}\right.---\left(3\right)$

So just adopt back two frame f _N+1, f _N+3Extrapolation present frame f _n

If SAD_num0, SAD_num1, SAD_num2 be greater than n, explanation exists camera lens to switch continuously so, at this moment no matter be interior insert or extrapolation all can not obtain good effect, to this situation, adopts the method for frame-skipping.So both guaranteed the quality of every width of cloth image, and for whole section video, picture is still very smooth.

Further; S104 comprises: with the coordinate of treating inserted block in the interpolation frame is benchmark, and in the hunting zone in two reference frames the position of mobile corresponding match block of symmetry, select minimum that of difference to piece as best matching blocks; To obtain motion vector, form the initial motion vectors field.

Usually adopt traditional estimation in the frame rate conversion; This method is to two known consecutive frames, is benchmark with a piece of a two field picture wherein, in another two field picture, searches for best matching blocks; Thereby obtain motion vector (MV), get the motion vector of MV/2 then as intermediate frame.This method can cause overlapping and the appearance leak phenomenon usually.

In the present invention, adopt symmetric motion estimation mode to carry out estimation, can solve the existing existing problem of conventional motion estimation that adopts.Symmetric motion estimates that be is benchmark with the coordinate of treating inserted block in the interpolation frame; The position of mobile corresponding match block of symmetry in certain hunting zone in two reference frames then; According to the matching criterior of formula (4), select minimum that of difference to piece as best matching blocks.Fig. 3 is the sketch map that symmetric motion is estimated.The estimation of this symmetry makes treats that each piece is " seamless " splicing in the contact pin, has therefore solved leak and overlap problem that the first motion estimation brings well.

$\mathrm{SAD}\_\mathrm{inter}\left[B_{\mathrm{ij}}\right]=\underset{s\&Element;B_{\mathrm{ij}}}{\mathrm{\&Sigma;}}\left|f_{n-1}\right[S-\mathrm{MV}\left]f_{n-1}\right[S+\mathrm{MV}\left]\right|---\left(4\right)$

But, the MV that obtains according to the matching criterior of formula (4) be in the video coding in order to improve code efficiency, make residual error MV hour, be not movement locus truly.

In FRUC, in order to make MV more near real movement locus, the correlation in the considered frame has simultaneously been revised the matching criterior of estimation, and use formula (5) is as matching criterior.

SAD[B _ij]＝SAD_inter[B _ij]+λSAD_intra[B _ij] (5)

$\mathrm{SAD}\_\mathrm{intra}\left[B_{\mathrm{ij}}\right]=\underset{h=0}{\overset{1}{\mathrm{\&Sigma;}}}\underset{l=0}{\overset{N-1}{\mathrm{\&Sigma;}}}|x_{\mathrm{kl}}-x_{\mathrm{kl}}^{\&prime;}|---\left(6\right)$

SAD_intra in the formula (6) representes the absolute difference of last horizontal boundary and left vertical boundary and their adjacent boundaries of current block, and N is one a width, and the size of piece is the size of N * N.λ is a weight coefficient, gets 0～16 integer.It is an empirical value, and for the more violent sequence of motion, the λ value should be less than normal, and to the milder sequence of motion, the λ value should be bigger than normal.

Further, S105 comprises:

The corresponding blocks that motion vector MV is labeled as L1 is divided into sub-piece;

With adjacent be labeled as L2 the drink piece is merged into bulk, to obtain variable-block;

Variable-block is carried out estimation, obtain new motion vector field.

Motion vector processing procedure in the face of the embodiment of the invention is elaborated down.

Symmetric motion estimates that the back just can obtain an initial MVF, and the most of MV in this MVF the inside is reliably, but still has the insecure MV of part.Insecure MV can cause corresponding blocks when interleave, to make a mistake, and then has a strong impact on subjective effect.Therefore, motion vector is handled very important part in MCFI.

In general, SAD more just matching error is bigger, and the MV that such piece is corresponding is insecure often, and therefore, SAD judges a whether reliable basic principle of MV.Though but very little for some piece SAD, its MV is also unreliable.The reason that this situation produces is that zone much similar or that repeat is arranged in the sub-picture, the corresponding several best matching blocks of possibility, and minimum according to SAD, the MV reliability that obtains is difficult to assurance.Such MV can judge and then handles through the correlation of MV.

In embodiments of the present invention, come the reliability of MV is classified in conjunction with the correlation of SAD and MV.

If (w h) is the position of current block in all pieces of piece image, and w is the capable piece number at current block place, and h is the row piece number at current place, (x _{W, h}, y _{W, h}) be the motion vector of current block.

Coefficient correlation between 8 adjacent blocks of current block around with it is defined as

$C_{w,h}=\frac{\frac{1}{8}\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}\underset{j=-1}{\overset{1}{\mathrm{\&Sigma;}}}\sqrt{{(x_{w,h}-x_{w+i,h+j})}^2+{(y_{w,h}-y_{w+i,h+j})}^2}}{\frac{1}{9}\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}\underset{j=-1}{\overset{1}{\mathrm{\&Sigma;}}}\sqrt{x_{w+i,h+j}^2+y_{w+i,h+j}^2}}---\left(7\right)$

Current block is defined as with the average correlation coefficient of adjacent block on every side

$C_{w,h}^{\mathrm{avg}}=\frac{1}{9}\underset{i=-1}{\overset{1}{\mathrm{\&Sigma;}}}\underset{j=-1}{\overset{1}{\mathrm{\&Sigma;}}}{C}_{w+i,h+j}---\left(8\right)$

All motion vectors of piece image are divided into 3 types, and the 1st type is because the excessive unreliable MV that causes of SAD is labeled as L1; The 2nd type is because the MV of current block and the unreliable MV that the MV correlation of piece does not on every side cause are labeled as L2 very much; Remaining the 3rd type is reliable MV, is labeled as L3.

Following with the above-mentioned motion vector classification of formula (9) expression:

SAD in the formula (9) is that formula (5) is obtained next result.

After motion vector classification, to take diverse ways to handle respectively for the unreliable MV of label L 1 and label L 2.The processing procedure of motion vector is following:

1, the corresponding blocks that MV is labeled as L1 is divided into sub-piece.Because such piece is the edge of object often, so it possibly comprise two different objects, therefore piece is divided into fritter, each piece comprises an object, can make the edge of object more clear like this;

2, with adjacent be labeled as L2 the drink piece is merged into bulk.Because the MV of L2 class appears at the very level and smooth zone of motion vector on every side often, and this MV that only is labeled is irregular, therefore to be merged into a bulk with piece on every side;

3, through after the cutting apart and merge of front, the size of piece has been not initial fixed size, has become the variable-block size.To these variable-blocks estimation again, obtain a new MVF;

4, the reliability to the MV after upgrading reclassifies;

5, repeat 4 above processes, reach the block size of 2*2 up to the SAD of L1 classification corresponding blocks less than a certain fixed threshold or smallest partition piece.

Further, S106 can comprise: adopt forward frame or consequent frame that present frame is carried out interpolation, perhaps use mode that bidirectional frame fills up that the motion vector of new motion vector field is carried out adaptive motion compensated processing, obtain to treat interleave.

Describe in the face of the adaptive motion compensated mode that the embodiment of the invention adopted down.

MCFI is actually the process of a linear filtering, and interpolation result can be represented with following formula:

${\hat{f}}_n[s,v]={\mathrm{\&lambda;f}}_{n-1}[s-\mathrm{MV}]+(1-\mathrm{\&lambda;})f_{n+1}[s+\mathrm{MV}]---\left(10\right)$

F in the formula _N-1With f _N+1Be two known consecutive frames,

It is the frame that prediction is come out.S representes a block, and MV is the corresponding motion vector of this piece.Weights λ generally gets 0.5.But the filter of this fixedly weights tends to let local fuzzy.The present invention selects λ by following formula:

If the SAD of current block has surpassed a certain threshold value, adopt forward frame or consequent frame that present frame is carried out interpolation so, otherwise, use bidirectional frame to fill up.

In embodiments of the present invention, taked diverse ways to promote the subjective effect of interpolation image respectively 3 of frame rate conversion different stages.At first, utilize scene detection to judge and be to use the method for interior method of inserting or extrapolation that original video sequence is handled, verification is upgraded in the initial motion vectors field that then estimation is obtained, to obtain motion vector (MV) the most reliably.At last select the weights of interpolation filter further to eliminate local blooming adaptively.Scene detection adopts frame difference method, and is simple and practical in frame rate conversion is used.Because it only needs the number of statistics SAD above the size of a certain threshold value, can reduce the complexity of computation complexity and estimation, little to the influence of overall complexity, in addition, can not only improve the quality of interpolation frame, also satisfy the requirement of using in real time simultaneously.

Fig. 4 is that the structure based on conversion equipment on the frame per second of motion vector processing of the embodiment of the invention is formed sketch map, and as shown in Figure 4, this device comprises:

Scene detection module 40 is used for the original video sequence of input is carried out scene detection, detects the current scene change generation that whether has;

Video sequence processing module 41 is used for then adopting the data after interpolation method is handled original video sequence and obtained to handle if scene detection module 40 detects scene change when taking place; If when scene detection module 40 does not detect the scene change generation, then adopt the data after the extrapolation mode is handled original video sequence and obtained to handle;

Motion vector acquisition module 42 is used to adopt symmetric motion to estimate the data after 41 processing of video sequence processing module are calculated, and obtains the initial motion vectors field;

Motion vector processing module 43, the motion vector of the initial motion vectors field that is used for motion vector acquisition module 42 is obtained are carried out motion vector and are handled and obtain new motion vector field;

Adaptive interpolation filters 44, the motion vector of the new motion vector field that is used for motion vector processing module 43 is obtained carries out adaptive motion compensated processing, obtains to treat interleave.

Whether in the practical implementation, scene detection module 40 also is used to adopt frame difference method to detect scene change, have scene change to take place according to adding up absolute difference and the SAD number judgment above the piece of specific threshold.

Further; Motion vector acquisition module 42 also is used for treating that with interpolation frame the coordinate of inserted block is a benchmark; The position of the mobile corresponding match block of symmetry in the hunting zone in two reference frames; Select minimum that of difference to piece as best matching blocks, to obtain motion vector, form the initial motion vectors field.

Further, the corresponding blocks that motion vector processing module 43 also is used for motion vector MV is labeled as L1 is divided into sub-piece, with adjacent be labeled as L2 the drink piece is merged into bulk, to obtain variable-block, variable-block is carried out estimation, obtain new motion vector field.

Further, adaptive interpolation filters 44 also is used to adopt forward frame or consequent frame that present frame is carried out interpolation, perhaps uses mode that bidirectional frame fills up that the motion vector of new motion vector field is carried out adaptive motion compensated processing, obtains to treat interleave.

Implementation procedure and principle based on each functions of modules of conversion equipment on the frame per second of motion vector processing in the embodiment of the invention can be described based on the respective process among the embodiment of conversion method on the frame per second of motion vector processing referring to of the present invention, repeat no more here.

One of ordinary skill in the art will appreciate that all or part of step in the whole bag of tricks of the foregoing description is to instruct relevant hardware to accomplish through program; This program can be stored in the computer-readable recording medium; Storage medium can comprise: read-only memory (ROM; Read Only Memory), random access memory (RAM, Random Access Memory), disk or CD etc.

In addition; More than conversion method and device thereof on the frame per second that the embodiment of the invention provides handles based on motion vector had been carried out detailed introduction; Used concrete example among this paper principle of the present invention and execution mode are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that on embodiment and range of application, all can change, in sum, this description should not be construed as limitation of the present invention.

Claims (10)

1. conversion method on the frame per second of handling based on motion vector is characterized in that said method comprises:

Original video sequence to input carries out scene detection, detects the current scene change generation that whether has;

If when detecting the scene change generation, then adopt the data after interpolation method is handled and obtained to handle said original video sequence;

If when not detecting the scene change generation, then adopt the data after the extrapolation mode is handled and obtained to handle said original video sequence;

Adopt symmetric motion to estimate the data after handling are calculated, obtain the initial motion vectors field;

Motion vector to said initial motion vectors field carries out the new motion vector field of motion vector processing acquisition;

Motion vector to said new motion vector field carries out adaptive motion compensated processing, obtains to treat interleave.

2. conversion method on the frame per second of handling based on motion vector as claimed in claim 1 is characterized in that said original video sequence to input carries out scene detection, and whether detects currently has the step of scene change generation to comprise:

Adopt frame difference method to detect scene change, whether have scene change to take place according to adding up absolute difference and SAD number judgment above the piece of specific threshold.

3. conversion method on according to claim 1 or claim 2 the frame per second of handling based on motion vector is characterized in that said employing symmetric motion is estimated the data after handling are calculated, and the step that obtains the initial motion vectors field comprises:

With the coordinate of treating inserted block in the interpolation frame is benchmark, and in the hunting zone in two reference frames the position of mobile corresponding match block of symmetry, select minimum that of difference to piece as best matching blocks, to obtain motion vector, form the initial motion vectors field.

4. conversion method on the frame per second of handling based on motion vector as claimed in claim 3 is characterized in that, said motion vector to said initial motion vectors field carries out motion vector to be handled the step that obtains new motion vector field and comprise:

The corresponding blocks that motion vector MV is labeled as L1 is divided into sub-piece;

With adjacent be labeled as L2 the drink piece is merged into bulk, to obtain variable-block;

Said variable-block is carried out estimation, obtain new motion vector field.

5. as claimed in claim 4 based on conversion method on the frame per second of motion vector processing; It is characterized in that; Said motion vector to said new motion vector field carries out adaptive motion compensated processing; Acquisition treats that the step of interleave comprises: adopt forward frame or consequent frame that present frame is carried out interpolation, perhaps use mode that bidirectional frame fills up that the motion vector of said new motion vector field is carried out adaptive motion compensated processing, obtain to treat interleave.

6. conversion equipment on the frame per second of handling based on motion vector is characterized in that said device comprises:

Scene detection module is used for the original video sequence of input is carried out scene detection, detects the current scene change generation that whether has;

The video sequence processing module is used for then adopting the data after interpolation method is handled and obtained to handle said original video sequence if said scene detection module detects scene change when taking place; If when said scene detection module does not detect the scene change generation, then adopt the data after the extrapolation mode is handled and obtained to handle said original video sequence;

The motion vector acquisition module is used to adopt symmetric motion to estimate the data after the said video sequence processing module processing are calculated, and obtains the initial motion vectors field;

Motion vector processing module, the motion vector of the initial motion vectors field that is used for said motion vector acquisition module is obtained are carried out motion vector and are handled and obtain new motion vector field;

Adaptive interpolation filters, the motion vector of the new motion vector field that is used for said motion vector processing module is obtained carries out adaptive motion compensated processing, obtains to treat interleave.

7. as claimed in claim 6 based on conversion equipment on the frame per second of motion vector processing; It is characterized in that; Whether said scene detection module also is used to adopt frame difference method to detect scene change, have scene change to take place according to adding up absolute difference and the SAD number judgment above the piece of specific threshold.

8. like conversion equipment on claim 6 or the 7 described frame per second of handling based on motion vector; It is characterized in that; Said motion vector acquisition module also is used for treating that with interpolation frame the coordinate of inserted block is a benchmark, and in the hunting zone in two reference frames the position of the mobile corresponding match block of symmetry, select minimum that of difference to piece as best matching blocks; To obtain motion vector, form the initial motion vectors field.

9. as claimed in claim 8 based on conversion equipment on the frame per second of motion vector processing; It is characterized in that; The corresponding blocks that said motion vector processing module also is used for motion vector MV is labeled as L1 is divided into sub-piece, with adjacent be labeled as L2 the drink piece is merged into bulk, to obtain variable-block; Said variable-block is carried out estimation, obtain new motion vector field.

10. as claimed in claim 9 based on conversion equipment on the frame per second of motion vector processing; It is characterized in that; Said adaptive interpolation filters also is used to adopt forward frame or consequent frame that present frame is carried out interpolation; Perhaps use mode that bidirectional frame fills up that the motion vector of said new motion vector field is carried out adaptive motion compensated processing, obtain to treat interleave.

Images