CN102215386B - Video image block processing method and device - Google Patents
Video image block processing method and device Download PDFInfo
- Publication number
- CN102215386B CN102215386B CN 201010144630 CN201010144630A CN102215386B CN 102215386 B CN102215386 B CN 102215386B CN 201010144630 CN201010144630 CN 201010144630 CN 201010144630 A CN201010144630 A CN 201010144630A CN 102215386 B CN102215386 B CN 102215386B
- Authority
- CN
- China
- Prior art keywords
- motion vector
- estimated motion
- template
- matching error
- cost function
- 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
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention relates to a video image block processing method and a video image block processing device. The method comprises the following steps of: acquiring matching errors between a current template of a current block and reference templates corresponding to more than one estimated motion vector; correcting the matching errors between the reference templates corresponding to a part or all of the more than one estimated motion vector and the current template according to the credibility of the estimated motion vector; selecting an optimal matched template from the more than one reference template according to the corrected matching errors; and obtaining a predicted block of the current block according to the optimal matched template, and coding or decoding the current block. By the method and the device, matching similarities between the reference templates corresponding to incredible estimated motion vectors and the current template are reduced, so the probability that the selected optimal matched template is the reference template corresponding to a credible estimated motion vector is increased, and wobble effects are worsened.
Description
Technical field
The embodiment of the invention relates to the communications field, relates in particular to a kind of video image block processing method and device.
Background technology
In Video coding, the inter prediction technology of motion compensation has very important effect to the raising of code efficiency.In existing video standard, encoder adopts the method for motion search to derive movable information, this movable information can be motion vector (Motion Vector, be called for short: relevant information MV), for example: movable information is the motion vector difference with respect to the spatial motion vectors predicted value, movable information is encoded in the code stream, and decoder decodes movable information from code stream, then utilizes motion compensation technique to come reconstructed image.For example, in standard H.264/AVC, encoder adopts the method for estimating of nonstandardized technique to derive and obtains movable information, movable information is encoded in the code stream, decoder decodes movable information from code stream, in reference frame, be displaced to the band of position of motion vector indication, inter prediction is carried out as the prediction signal of current block in this band of position.In this inter prediction technology, movable information will take a large amount of coded bit streams.
In order to reduce coded bit stream, prior art proposes a kind of decoding end motion vector (Decoder Side Motion Vector Derivation that can derive, be called for short: DMVD) technology, it is applied in the interframe prediction coding and decoding, need in code stream, not transmit movable information.Particularly, at coding side, the reconstructed image that obtains the left side of current block and top consists of when front template (TM1), the reference template (TM2) of search and current template matches in the hunting zone of appointment in the reference frame of current block, and the shape of TM1 and TM2 is the same; When the matching error of TM1 and TM2 hour, the prediction piece adjacent with TM2 predicted the residual error that obtains current block as the prediction signal of current block to current block, carries out carrying out the entropy coding after transform and quantization processes again.In decoding end, utilize method search same as described above to obtain TM2, according to the prediction piece adjacent with TM2, rebuild current block.
Further, the method for fast searching of predictability has replaced above-mentioned all direction search method in the DMVD technology, the method is with the motion vector of encoded image piece around the current block set of candidate motion vectors as current block, because the room and time correlation of moving object, the DMVD technology can be selected the optimum motion vector as current block in these set of candidate motion vectors, and needn't as conventional motion vector all direction search method, search within the specific limits, and according to certain level from whole pixel, minute pixel is searched for step by step, thereby has significantly reduced the calculating pressure of decoding end.
Since need to be in code stream in the DMVD technology explicit encoding motion vector, adopt the DMVD technology can reduce a large amount of coded bit streams.If there is not the movable information coding, then can reduce the needed bit of coding.(Rate Distortion Optimization is called for short: RDO) under the mode adjudging, for the continuous motion image-region, often adopt the DMVD technology to encode in rate-distortion optimal.Do not have architectural feature or smoother if these zones are relative, then for each different motion vector candidates value, the template matches error amount of DMVD may become very close.
In the process that realizes the invention, the inventor finds: owing to have quantizing noise in template or the reference frame, DMVD may derive and obtain one with respect to the motion vector of the true nature dyskinesia in the video sequence.Accordingly, when obtain in time this or the space on during irregular motion vector, can cause affecting the jitter block effect of subjective vision effect.
Summary of the invention
The embodiment of the invention provides a kind of video image block processing method and device, in order to weaken the jitter block effect.
The embodiment of the invention provides a kind of video image block processing method, comprising:
Obtain the matching error when the front template reference template corresponding with above estimated motion vector of current block;
According to the confidence level of described estimated motion vector, partly or entirely estimated motion vector is corresponding in the described above estimated motion vector reference template and the described matching error of working as front template are revised;
According to revised matching error, from an above reference template, select the optimum Match template;
According to described optimum Match template, obtain the prediction piece of current block, described current block is encoded or decoded.
The embodiment of the invention provides a kind of video image blocks processing unit, comprising:
Matching module is used for obtaining the matching error when the front template reference template corresponding with above estimated motion vector of current block;
Correcting module is used for the confidence level according to described estimated motion vector, and partly or entirely estimated motion vector is corresponding in the described above estimated motion vector reference template and the described matching error of working as front template are revised;
Select module, be used for according to revised matching error, from an above reference template, select the optimum Match template;
Processing module is used for according to described optimum Match template, obtains the prediction piece of current block, and described current block is encoded or decoded.
The embodiment of the invention is revised matching error according to the confidence level of estimated motion vector, from an above reference template, select the optimum Match template according to revised matching error, according to the optimum Match template, find the prediction piece of current block, current block is encoded or decoded.By matching error is revised, can be so that reference template corresponding to incredible estimated motion vector be weakened with the matching similarity when front template, the optimum Match template of selecting like this is that the possibility of reference template corresponding to believable estimated motion vector increases, and has also just weakened flutter effect.
Description of drawings
In order to be illustrated more clearly in the technical scheme in the embodiment of the invention, the below will do to introduce simply to the accompanying drawing of required use among the embodiment, apparently, accompanying drawing in the following describes only is some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain according to these accompanying drawings other accompanying drawing.
The flow chart of the video image block processing method that Fig. 1 provides for the embodiment of the invention one;
The flow chart of the video image block processing method that Fig. 2 provides for the embodiment of the invention two;
The flow chart of the video image block processing method that Fig. 3 provides for the embodiment of the invention three;
The flow chart of the video image block processing method that Fig. 4 provides for the embodiment of the invention four;
The schematic diagram of present frame in the video image block processing method that Fig. 5 provides for the embodiment of the invention four;
Fig. 6 is the schematic diagram of reference frame corresponding to the present frame of Fig. 5;
The flow chart of the video image block processing method that Fig. 7 provides for the embodiment of the invention six;
The structural representation of the video image blocks processing unit that Fig. 8 provides for the embodiment of the invention seven;
The structural representation of the video image blocks processing unit that Fig. 9 provides for the embodiment of the invention eight;
The structural representation of the video image blocks processing unit that Figure 10 provides for the embodiment of the invention nine;
The structural representation of the video image blocks processing unit that Figure 11 provides for the embodiment of the invention ten;
The structural representation of the video image blocks processing unit that Figure 12 provides for the embodiment of the invention 11;
The structural representation of the video image blocks processing unit that Figure 13 provides for the embodiment of the invention 12.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is clearly and completely described, 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 belong to the scope of protection of the invention not making the every other embodiment that obtains under the creative work prerequisite.
The flow chart of the video image block processing method that Fig. 1 provides for the embodiment of the invention one.As shown in Figure 1, present embodiment comprises:
101, obtain the matching error when the front template reference template corresponding with above estimated motion vector of current block.
In the method for fast searching of predictability, utilize the reconstructed pixels structure of current block periphery to work as front template, with the motion vector of encoded image piece around the current block set of candidate motion vectors as current block, these candidate motion vectors are as the part of estimated motion vector; Peripheral region (as respectively being offset up and down the zone of a search precision unit) in the candidate motion vector indication band of position, the search that becomes more meticulous obtains some reference templates the same with current shape of template, and these reference templates are corresponding becomes more meticulous motion vector also as the part of estimated motion vector.
Obtain respectively the matching error of working as front template and above-mentioned candidate motion vector and reference template corresponding to motion vector that become more meticulous of current block.
102, according to the confidence level of estimated motion vector, partly or entirely estimated motion vector is corresponding in the above estimated motion vector reference template and the matching error when front template are revised.
The confidence level of so-called estimated motion vector refers to estimate the motion vector that obtains and the consistency of the real motion of this object, a part is considered to believable (being that confidence level is higher) in the above estimated motion vector, another part is considered to incredible (being that confidence level is lower), for example: because the spatial continuity of object of which movement, it is believable that candidate motion vector can be considered to, and the motion vector that becomes more meticulous is considered to incredible.
In the present embodiment, can revise with the matching error when front template reference template corresponding to believable estimated motion vector, also can revise with the matching error when front template reference template corresponding to incredible estimated motion vector, can also revise with the matching error when front template by corresponding reference template whole estimated motion vectors.
103, according to revised matching error, from an above reference template, select the optimum Match template.
After revising, the size of comparison match error can select the reference template of matching error minimum as the optimum Match template.
104, according to the optimum Match template, obtain the prediction piece of current block, described current block is encoded or decoded.
According to the optimum Match template, by motion vector corresponding to this optimum Match template, find the prediction piece of current block, described current block is encoded or decoded.
Present embodiment is revised matching error according to the confidence level of estimated motion vector, from an above reference template, select the optimum Match template according to revised matching error, according to the optimum Match template, find the prediction piece of current block, current block is encoded or decoded.By matching error is revised, can be so that reference template corresponding to incredible estimated motion vector be weakened with the matching similarity when front template, the optimum Match template of selecting like this is that the possibility of reference template corresponding to believable estimated motion vector increases, and has also just weakened flutter effect.
The flow chart of the video image block processing method that Fig. 2 provides for the embodiment of the invention two.As shown in Figure 2, present embodiment comprises:
201, obtain the matching error when the front template reference template corresponding with above estimated motion vector of current block.
The specific implementation process of this step can be identical with step 101.
202, the reference template that the part estimated motion vector is corresponding and the cost function of presetting when the first matching error substitution of front template obtain the first cost function value as revised matching error.
Confidence level according to estimated motion vector, an above estimated motion vector can be divided into two parts, a part is believable motion vector, a part is incredible motion vector, the described part estimated motion vector of this step is incredible motion vector, incredible motion vector with a low credibility in believable motion vector.The reference template that incredible motion vector is corresponding and the cost function of presetting when the first matching error substitution of front template, obtain the first cost function value as revised matching error, wherein the first cost function value is greater than the first matching error, that is to say that the matching error of incredible motion vector had increased after process was revised.
As a kind of embodiment, default cost function can for:
C＝ω·SAD(MV) (1)
Wherein, SAD (MV) is the first matching error, be specially absolute error and, can certainly for the difference of two squares and or square root and, can also be the functions relevant with distortion such as code rate distortion, present embodiment does not limit this; ω is the punishment parameter, and the first cost function value C that obtains after ω＞1, the first matching error SAD (MV) process is revised here is greater than SAD (MV).
Cost function shown in the formula (1) only is a concrete example, and present embodiment is not limited only to this, all belongs to the claimed scope of present embodiment so long as guarantee revised the first cost function value greater than the cost function of the first matching error.
203, select minimum one in the revised matching error of part with in through the matching error of revising, the reference template that it is corresponding is as the optimum Match template.
Present embodiment is revised the matching error of incredible motion vector, the matching error of believable motion vector is not through revising, from these revised matching errors and through selecting minimum one the matching error of revising, reference template corresponding to the matching error that this is minimum is as the optimum Match template.
204, according to the optimum Match template, obtain the prediction piece of current block, described current block is encoded or decoded.
According to the optimum Match template, by motion vector corresponding to this optimum Match template, find the prediction piece of current block, described current block is encoded or decoded.
Present embodiment is revised the matching error of incredible motion vector, revised matching error has increased, so that reference template corresponding to incredible estimated motion vector is weakened with the matching similarity when front template, then select reference template corresponding to minimum matching error as the optimum Match template, the optimum Match template of selecting is that the possibility of reference template corresponding to believable estimated motion vector increases, and has also just weakened flutter effect.
The flow chart of the video image block processing method that Fig. 3 provides for the embodiment of the invention three.As shown in Figure 3, present embodiment comprises:
301, obtain the matching error when the front template reference template corresponding with above estimated motion vector of current block.
The specific implementation process of this step can be identical with step 101.
302, the reference template that the part estimated motion vector is corresponding and the cost function of presetting when the second matching error substitution of front template obtain the second cost function value as revised matching error.
Confidence level according to estimated motion vector, an above estimated motion vector can be divided into two parts, a part is believable motion vector, a part is incredible motion vector, the described part estimated motion vector of this step is believable motion vector, believable motion vector with a high credibility in incredible motion vector.The reference template that believable motion vector is corresponding and the cost function of presetting when the second matching error substitution of front template, obtain the second cost function value as revised matching error, wherein the second cost function value is less than the second matching error, that is to say that the matching error of believable motion vector had reduced after process was revised.
As a kind of embodiment, default cost function can for:
C＝ω·SAD(MV) (2)
Wherein, SAD (MV) is the second matching error, be specially absolute error and, can certainly for the difference of two squares and or square root and, can also be the functions relevant with distortion such as code rate distortion, present embodiment does not limit this; ω is the punishment parameter, and the second cost function value C that obtains after ω＜1, the second matching error SAD (MV) process is revised here is less than SAD (MV).
Cost function shown in the formula (2) only is a concrete example, and present embodiment is not limited only to this, all belongs to the claimed scope of present embodiment so long as guarantee revised the second cost function value less than the cost function of the second matching error.
303, select minimum one in the revised matching error of part with in through the matching error of revising, the reference template that it is corresponding is as the optimum Match template.
Present embodiment is revised the matching error of believable motion vector, the matching error of incredible motion vector is not through revising, from these revised matching errors and through selecting minimum one the matching error of revising, reference template corresponding to the matching error that this is minimum is as the optimum Match template.
304, according to the optimum Match template, obtain the prediction piece of current block, described current block is encoded or decoded.
According to the optimum Match template, by motion vector corresponding to this optimum Match template, find the prediction piece of current block, described current block is encoded or decoded.
Present embodiment is revised the matching error of believable motion vector, revised matching error has reduced, so that reference template corresponding to believable estimated motion vector is enhanced with the matching similarity of working as front template, then select reference template corresponding to minimum matching error as the optimum Match template, the optimum Match template of selecting is that the possibility of reference template corresponding to believable estimated motion vector increases, and has also just weakened flutter effect.
The flow chart of the video image block processing method that Fig. 4 provides for the embodiment of the invention four.As shown in Figure 4, present embodiment comprises:
401, obtain the matching error when the front template reference template corresponding with above estimated motion vector of current block.
In the method for fast searching of predictability, with the motion vector of encoded image piece around the current block set of candidate motion vectors as current block, candidate motion vector is known motion vector or the whole pixel precision motion vector of current block periphery, and these candidate motion vectors are as the part of estimated motion vector; Peripheral region (as respectively being offset up and down the zone of a search precision unit) in the candidate motion vector indication band of position, the search that becomes more meticulous obtains some reference templates, the motion vector that becomes more meticulous that these reference templates are corresponding is also as the part of estimated motion vector, and the motion vector that becomes more meticulous is motion vector or the half-pixel accuracy motion vector that candidate motion vector obtains by skew half-pixel accuracy position.Describe below by a concrete example:
The schematic diagram of present frame in the video image block processing method that Fig. 5 provides for the embodiment of the invention four, Fig. 6 is the schematic diagram of reference frame corresponding to the present frame of Fig. 5.Wherein, present frame refers to the frame under the image block (being current block) of pre-treatment.As shown in Figure 5, the encoded image piece comprises left side piece A, upper right C and upper left C ' around the current block, the set of candidate motion vectors { mv1, mv2, the mv3 that their motion vector are formed current block, mv4}, mv1 wherein, mv2, mv3 are respectively the motion vector of left side piece A, upper right C and upper left C ', mv4 is mv1, mv2 and mv3 mean value.The peripheral region of each candidate motion vector indication band of position in reference frame, the search that becomes more meticulous obtains some reference templates, and these reference templates are corresponding becomes more meticulous motion vector also as the part of estimated motion vector.As shown in Figure 6, take upper left C ' as example, position in reference frame shown in the circle is the motion vector mv3 indication band of position of upper left C ', carry out the sub-pix displacement in this band of position, obtain the deviation post of 4 triangle indications, the reference template that forms at the deviation post of these 4 triangle indications is the reference template that the search that becomes more meticulous obtains, these reference template correspondences motion vector { mv31 that becomes more meticulous, mv32, mv33, mv34}, these become more meticulous motion vector and set of candidate motion vectors consist of the estimated motion vector collection jointly.
This step is obtained respectively the matching error when front template and above-mentioned candidate motion vector and reference template corresponding to motion vector that become more meticulous of current block, when the shape of front template and reference template is the same.Wherein matching error can be absolute error and (Sum of Absolute Differences, be called for short: SAD), also can for the difference of two squares and (Sum of Squares for Error, be called for short: SSE) or square root and, can also be the functions relevant with distortion such as code rate distortion.
402, the cost function that the matching error substitution is default according to the character of estimated motion vector, obtains cost function value as revised matching error.
For the continuous motion image-region, if there is not architectural feature or smoother, the value of matching error may be very close, the in addition impact of quantizing noise is directly selected the optimum Match template to derive according to matching error and is obtained one with respect to the motion vector of the true nature dyskinesia in the video sequence.This step weakens the jitter block effect under the prerequisite that guarantees than high coding efficiency, and the method that provides a kind of weakening to become more meticulous the template matches similarity particularly, utilizes some conditional-variables to construct cost function, is used for revising matching error.
Present embodiment provides a kind of cost function, and the parameter of this cost function comprises the punishment parameter, and the value of punishment parameter is according to the confidence level of estimated motion vector and definite.Further, in the present embodiment, whether the confidence level of estimated motion vector is again to be the candidate motion vector of current block and definite according to estimated motion vector.Think the with a high credibility in the motion vector that becomes more meticulous of candidate motion vector in the present embodiment, so the value of punishment parameter corresponding to candidate motion vector is less than the value of punishment parameter corresponding to the motion vector that becomes more meticulous.
As a kind of embodiment, default cost function can for:
C＝ω·SAD(MV)，
Wherein, ω for punishment parameter, SAD (MV) be matching error, be specially absolute error and, can certainly for the difference of two squares and or square root with, can also be the functions relevant with distortion such as code rate distortion, present embodiment does not limit this; C is cost function value.When estimated motion vector was original candidate motion vector, the value of punishment parameter was 1, and when estimated motion vector is not original candidate motion vector, but when becoming more meticulous motion vector, the value of punishment parameter is
Wherein
The value that also is the punishment parameter of estimated motion vector when being candidate motion vector is the value of the punishment parameter when becoming more meticulous motion vector less than estimated motion vector, like this according to formula (3), for the cost function value of reference template corresponding to the motion vector that the becomes more meticulous cost function value greater than reference template corresponding to candidate motion vector, thereby so that reference template corresponding to the motion vector that becomes more meticulous is weakened with respect to candidate motion vector with the matching similarity when front template.
Can comprise the punishment parameter in the code stream, the value of punishment parameter is predefined at coding side, concrete value can be set by off-line training, after coding side is according to default punishment calculation of parameter cost function value, can will should punish in the parameter read-in code stream like this; When decoding end is decoded, can from code stream, decode and obtain punishing parameter, according to this punishment calculation of parameter cost function value.Also can not comprise the punishment parameter in the code stream, the value of punishment parameter is predefined with decoding end and be identical at coding side, concrete value can be set by off-line training, and coding side uses respectively predefined identical punishment calculation of parameter cost function value with decoding end.
As another embodiment, default cost function can for:
C＝ω·SAD(MV)+δ，
Wherein, ω for punishment parameter, SAD (MV) be matching error, be specially absolute error and, can certainly for the difference of two squares and or square root with, can also be the functions relevant with distortion such as code rate distortion, present embodiment does not limit this; C is cost function value, and δ is the correction value of presetting, and its value can obtain by off-line training.When estimated motion vector was original candidate motion vector, the value of punishment parameter was 1, and when estimated motion vector is not original candidate motion vector, but when becoming more meticulous motion vector, the value of punishment parameter is
Wherein
The value that also is the punishment parameter of estimated motion vector when being candidate motion vector is the value of the punishment parameter when becoming more meticulous motion vector less than estimated motion vector, like this according to formula (4), the cost function value of the reference template that the motion vector that becomes more meticulous is corresponding is greater than the cost function value of reference template corresponding to candidate motion vector, thereby so that reference template corresponding to the motion vector that becomes more meticulous is weakened with respect to candidate motion vector with the matching similarity when front template.
From the above, in the present embodiment cost function determine be subject to matching error and the estimated motion vector corresponding with reference template is the impact of original candidate motion vector or the motion vector that becomes more meticulous.
Cost function shown in formula (3) and (4) only is a concrete example; present embodiment is not limited only to this, so long as reference template corresponding to the motion vector that becomes more meticulous after guarantee revising with all belong to the claimed scope of present embodiment when the matching similarity of the front template cost function weakened with respect to candidate motion vector.
403, from an above reference template, select revised matching error minimum as the optimum Match template.
The cost function value that relatively calculates selects the reference template of cost function minimum as the optimum Match template.
404, the image block of the estimated motion vector indication that the optimum Match template is corresponding is encoded or is decoded current block as the prediction piece.
At coding side, the image block of the estimated motion vector indication that this optimum Match template is corresponding is predicted the residual error that obtains current block as the prediction piece to current block, carries out carrying out the entropy coding after transform and quantization processes again; In decoding end, the image block of the estimated motion vector indication that the optimum Match template is corresponding is rebuild current block as the prediction piece according to this prediction piece.
In the present embodiment, directly the above-mentioned cost function of transmission in code stream after coding side is according to the cost function calculation cost function value, can write cost function in the code stream; When decoding end is decoded, can from code stream, decode and obtain cost function, according to this cost function calculation cost function value.The index value that also can comprise cost function in the code stream, in this case, the cost function that might adopt for different current blocks is different, adopt the cost function of formula (3) such as some current blocks, other current blocks adopt the cost function of formula (4), the index value of transmission cost function in the code stream, encoding and decoding end select corresponding cost function to calculate according to index value respectively.
Present embodiment is by the cost function of structure, matching error correction is obtained cost function value, the revised matching error of the reference template that motion vector is corresponding so that become more meticulous is larger, the revised matching error of the reference template that candidate motion vector is corresponding is less, then from an above reference template, select the optimum Match template according to revised matching error, according to the optimum Match template, find the prediction piece of current block, current block is encoded or decoded, the reference template that the search that becomes more meticulous in the present embodiment obtains is weakened with the matching similarity confidence level when front template, therefore the optimum Match template of selecting is that the possibility of reference template corresponding to candidate motion vector increases, and has so also just weakened the flutter effect that brings because of the search that becomes more meticulous.
The video image block processing method that the embodiment of the invention five provides is that with the difference of above-described embodiment four confidence level of estimated motion vector is definite with the texture structure information of working as front template according to reference template corresponding to estimated motion vector, also be that the value of punishment parameter in the cost function is subject to matching error and when the impact of the texture structure information of front template and reference template, further, the confidence level of estimated motion vector is definite with the similarity degree of the variance of working as front template or gradient information according to reference template corresponding to estimated motion vector.
Take variance as example, present embodiment is on the basis of above-described embodiment four, replace with following method and to determine cost function according to the character of estimated motion vector: the variance of working as front template (representing with the TM1) reference template (with TM2 represent) corresponding with above estimated motion vector of obtaining current block, wherein the variance of TM1 is the D1=∑ | TM1 (x, y)-E1|, wherein E1 is the average of each pixel value of TM1, TM1 (x, y) be that TM1 is at the pixel value of (x, y) position; The variance of TM2 is the D2=∑ | TM2 (x, y)-E2|, and wherein E2 is the average of each pixel value of TM2, TM2 (x, y) is that TM2 is at the pixel value of (x, y) position.
As a kind of embodiment, default cost function can for:
C＝ω·SAD(MV)，
Wherein, ω for punishment parameter, SAD (MV) be matching error, be specially absolute error and, can certainly for the difference of two squares and or square root with, can also be the functions relevant with distortion such as code rate distortion, present embodiment does not limit this; C is cost function value.When D1/D2＜1.5 and D1/D2＞2/3, the value of punishment parameter is 1, and when D1/D2＜2/3 or D1/D2＞1.5, the value of punishment parameter is
Wherein
Also be, the value of punishment parameter is according to when the texture similarity degree of front template and reference template and definite, when D1/D2＜2/3 or D1/D2＞1.5, show that the texture structure when front template and reference template differs larger, the estimated motion vector that this reference template is corresponding is considered to insincere motion vector, namely think with a low credibility, the value of punishment parameter is
When D1/D2＜1.5 and D1/D2＞2/3, show when the texture structure of front template and reference template more similarly, the estimated motion vector that this reference template is corresponding is considered to credible motion vector, and is namely with a high credibility, and the value of punishment parameter is 1.
As another embodiment, default cost function can for:
C＝ω·SAD(MV)+δ，
Wherein, ω for punishment parameter, SAD (MV) be matching error, be specially absolute error and, can certainly for the difference of two squares and or square root with, can also be the functions relevant with distortion such as code rate distortion, present embodiment does not limit this; C is cost function value, and δ is the correction value of presetting, and its value can obtain by off-line training.When D1/D2＜1.5 and D1/D2＞2/3, the value of punishment parameter is 1, and when D1/D2＜2/3 or D1/D2＞1.5, the value of punishment parameter is
Wherein
Also be, the value of punishment parameter is according to when the texture similarity degree of front template and reference template and definite, when D1/D2＜2/3 or D1/D2＞1.5, show that the texture structure when front template and reference template differs larger, the estimated motion vector that this reference template is corresponding is considered to insincere motion vector, and the value of punishment parameter is
When D1/D2＜1.5 or D1/D2＞2/3, show when the texture structure of front template and reference template more similarly, the estimated motion vector that this reference template is corresponding is considered to credible motion vector, and the value of punishment parameter is 1.
Present embodiment also can be determined according to the similarity degree of the gradient information of working as front template and reference template the value of punishment parameter.
Because when the texture structure information of front template and reference template also is to weigh the image block of motion vector indication corresponding to reference template as a kind of method of current block predicted value confidence level, so above-mentioned value and the candidate motion vector that according to estimated motion vector whether is current block by determining the punishment parameter when the texture structure information of front template and reference template determines to punish that the effect of the value of parameter is identical.
The cost function of present embodiment structure obtains cost function value according to texture structure information to matching error correction, so that reference template and the revised matching error of reference template that differs greatly when the front template texture are larger, reference template and when the less revised matching error of reference template of front template texture difference less, then from an above reference template, select the optimum Match template according to revised matching error, according to the optimum Match template, find the prediction piece of current block, current block is encoded or decoded, the reference template that the texture diversity ratio is larger in the present embodiment is weakened with the matching similarity when front template, therefore the optimum Match template of selecting is that the possibility of the less reference template of texture difference increases, and has so also just weakened the flutter effect that brings because falsely dropping non-true motion vector.
The flow chart of the video image block processing method that Fig. 7 provides for the embodiment of the invention six.As shown in Figure 7, present embodiment comprises:
501, obtain the matching error when the front template reference template corresponding with above estimated motion vector of current block.
502, the cost function that the matching error substitution is default according to the character of estimated motion vector, obtains cost function value as revised matching error.
503, from an above reference template, select revised matching error minimum as the optimum Match template.
If 504 present embodiments are applied to coding side, this step is specially: when estimated motion vector corresponding to optimum Match template is the first estimated motion vector, adopt the DMVD technology that current block is encoded; When estimated motion vector corresponding to optimum Match template is the second estimated motion vector, movable information is write code stream, adopt non-DMVD technology that current block is encoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector;
If present embodiment is applied to decoding end, this step is specially: when estimated motion vector corresponding to optimum Match template is the first estimated motion vector, adopt the DMVD technology that current block is decoded; When estimated motion vector corresponding to optimum Match template is the second estimated motion vector, according to the movable information in the code stream, adopt non-DMVD technology that current block is decoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector.
The above-mentioned movable information that writes code stream can be set for the motion vector and that obtains by prior art estimation H.264/AVC the difference of motion vector.
Specifically, the description of the specific implementation process of above-mentioned steps 501-503 can referring to the step of embodiment two, three, four or five correspondences, not repeat them here.
In above-mentioned steps 504, if present embodiment is applied to coding side, when estimated motion vector corresponding to optimum Match template is believable estimated motion vector, adopt the DMVD technology that current block is encoded, also be about to the image block of estimated motion vector indication corresponding to optimum Match template as the prediction piece, current block is encoded; When estimated motion vector corresponding to optimum Match template is incredible estimated motion vector, adopt non-DMVD technology (such as the technology in H.264/AVC) that current block is encoded.If present embodiment is applied to decoding end, when estimated motion vector corresponding to optimum Match template is believable estimated motion vector, adopt the DMVD technology that current block is decoded, also be about to the image block of estimated motion vector indication corresponding to optimum Match template as the prediction piece, rebuild current block according to this prediction piece; When estimated motion vector corresponding to optimum Match template is incredible estimated motion vector, according to the movable information in the code stream, adopt non-DMVD technology (such as the technology in H.264/AVC) that current block is decoded.
When estimated motion vector corresponding to optimum Match template was believable estimated motion vector, present embodiment adopted the DMVD technology of the method for fast searching of predictability to encode or decode, and has weakened the jitter block effect.When estimated motion vector corresponding to optimum Match template is incredible estimated motion vector, if adopting in this case the DMVD technology encodes or decodes, some shakes can occur again, present embodiment coding side in this case, adopt non-DMVD technology to encode, decoding end is according to the movable information in the code stream (such as motion vector difference), adopt non-DMVD technology to decode, since in the non-DMVD technology according to the RDO criterion, in the destructuring zone, coding side more is ready to select less motion vector difference to encode, and can produce so the motion vector field of rule, guarantee the motion vector field regularization, also weakened the jitter block effect.
The structural representation of the video image blocks processing unit that Fig. 8 provides for the embodiment of the invention seven.As shown in Figure 8, present embodiment comprises: matching module 11, correcting module 12, selection module 13 and processing module 14, wherein:
The confidence level that correcting module 12 is used for according to estimated motion vector is revised partly or entirely estimated motion vector is corresponding in the above estimated motion vector reference template and the matching error when front template;
Specifically, in the method for fast searching of predictability, with the motion vector of encoded image piece around the current block set of candidate motion vectors as current block, these candidate motion vectors are as the part of estimated motion vector; Peripheral region (as respectively being offset up and down the zone of a search precision unit) in the candidate motion vector indication band of position, the search that becomes more meticulous obtains some reference templates, and these reference templates are corresponding becomes more meticulous motion vector also as the part of estimated motion vector.Matching module 11 obtains respectively the matching error of working as front template and above-mentioned candidate motion vector and reference template corresponding to motion vector that become more meticulous of current block.
A part is considered to believable (being that confidence level is higher) in the above estimated motion vector, another part is considered to incredible (being that confidence level is lower), for example: candidate motion vector can be considered to believable, and the motion vector that becomes more meticulous is considered to incredible.In the present embodiment, correcting module 12 can be revised with the matching error when front template reference template corresponding to believable estimated motion vector, also can revise with the matching error when front template reference template corresponding to incredible estimated motion vector, can also revise with the matching error when front template by corresponding reference template whole estimated motion vectors.
After revising, select the size of module 13 comparison match errors, can select the reference template of matching error minimum as the optimum Match template.
In the present embodiment, correcting module is revised matching error according to the confidence level of estimated motion vector, select module from an above reference template, to select the optimum Match template according to revised matching error, processing module is according to the optimum Match template, find the prediction piece of current block, current block is encoded or decoded.By matching error is revised, can be so that reference template corresponding to incredible estimated motion vector be relatively weakened with the matching similarity when front template, the optimum Match template of selecting like this is that the possibility of reference template corresponding to believable estimated motion vector relatively increases, and has also just weakened flutter effect.
The structural representation of the video image blocks processing unit that Fig. 9 provides for the embodiment of the invention eight.As shown in Figure 9, present embodiment is on the basis of above-described embodiment seven, further, correcting module 12 comprises: the first amending unit 21, be used for the reference template that the part estimated motion vector is corresponding and the cost function of presetting when the first matching error substitution of front template, obtain the first cost function value as revised matching error.
Above-mentioned part estimated motion vector with a low credibility in other estimated motion vectors except the part estimated motion vector, the first cost function value is greater than the first matching error.
Confidence level according to estimated motion vector, an above estimated motion vector can be divided into two parts, a part is believable motion vector, a part is incredible motion vector, the part estimated motion vector is incredible motion vector in the present embodiment, incredible motion vector with a low credibility in believable motion vector.Reference template and the cost function of presetting when the first matching error substitution of front template that the first amending unit 21 is corresponding with incredible motion vector, obtain the first cost function value as revised matching error, wherein the first cost function value is greater than the first matching error, that is to say that the matching error of incredible motion vector had increased after process was revised.
Further, the cost function that adopts of present embodiment can be formula (1).
The implementation procedure of each module concrete function of video image blocks processing unit that present embodiment provides can be with reference to the associated description among the said method embodiment two.
Present embodiment the first amending unit is revised the matching error of incredible motion vector, revised matching error has increased, so that reference template corresponding to incredible estimated motion vector is weakened with the matching similarity when front template, then select reference template corresponding to minimum matching error as the optimum Match template, the optimum Match template of selecting is that the possibility of reference template corresponding to believable estimated motion vector increases, and has also just weakened flutter effect.
The structural representation of the video image blocks processing unit that Figure 10 provides for the embodiment of the invention nine.As shown in figure 10, present embodiment is on the basis of above-described embodiment seven, further, correcting module 12 comprises: the second amending unit 31, be used for the reference template that the part estimated motion vector is corresponding and the cost function of presetting when the second matching error substitution of front template, obtain the second cost function value as revised matching error.
Above-mentioned part estimated motion vector with a high credibility in other estimated motion vectors except the part estimated motion vector, the second cost function value is less than the second matching error.
Confidence level according to estimated motion vector, an above estimated motion vector can be divided into two parts, a part is believable motion vector, a part is incredible motion vector, the part estimated motion vector is believable motion vector in the present embodiment, believable motion vector with a high credibility in incredible motion vector.Reference template and the cost function of presetting when the second matching error substitution of front template that the second amending unit 31 is corresponding with believable motion vector, obtain the second cost function value as revised matching error, wherein the second cost function value is less than the second matching error, that is to say that the matching error of believable motion vector had reduced after process was revised.
Further, the cost function that adopts of present embodiment can be formula (2).
The implementation procedure of each module concrete function of video image blocks processing unit that present embodiment provides can be with reference to the associated description among the said method embodiment three.
Present embodiment the second amending unit is revised the matching error of believable motion vector, revised matching error has reduced, so that reference template corresponding to believable estimated motion vector is enhanced with the matching similarity of working as front template, relatively, the reference template that incredible estimated motion vector is corresponding is weakened with the matching similarity when front template, then select reference template corresponding to minimum matching error as the optimum Match template, the optimum Match template of selecting is that the possibility of reference template corresponding to believable estimated motion vector increases, and has also just weakened flutter effect.
The structural representation of the video image blocks processing unit that Figure 11 provides for the embodiment of the invention ten.As shown in figure 11, present embodiment is on the basis of above-described embodiment seven, further, correcting module 12 comprises: the 3rd amending unit 41, be used for a reference template and the cost function of presetting when the matching error substitution of front template that above estimated motion vector is corresponding is obtained cost function value as revised matching error.
In the present embodiment, the parameter of cost function comprises the punishment parameter, and the value of punishment parameter is according to the confidence level of estimated motion vector and definite.Further, whether the confidence level of estimated motion vector can be to be the candidate motion vector of current block and definite according to estimated motion vector, think like this with a high credibility in the motion vector that becomes more meticulous of candidate motion vector, the value of the punishment parameter that candidate motion vector is corresponding is less than the value of punishment parameter corresponding to the motion vector that becomes more meticulous.The confidence level of estimated motion vector also can be definite with the texture structure information of working as front template according to reference template corresponding to estimated motion vector, further, the confidence level of estimated motion vector is definite with the similarity degree of the variance of working as front template or gradient information according to reference template corresponding to estimated motion vector.
Further, the cost function that adopts of present embodiment can be in formula (3)-(6) any one.
The implementation procedure of each module concrete function of video image blocks processing unit that present embodiment provides can be with reference to the associated description among said method embodiment four or five.
Present embodiment the 3rd amending unit is revised the matching error of whole estimated motion vectors, wherein the value of punishment parameter is to determine according to the confidence level of estimated motion vector in the cost function, like this so that reference template corresponding to incredible estimated motion vector is weakened with the matching similarity when front template, then select reference template corresponding to minimum matching error as the optimum Match template, the optimum Match template of selecting is that the possibility of reference template corresponding to believable estimated motion vector increases, and has also just weakened flutter effect.
On the basis of above-described embodiment seven to ten any embodiment, as a kind of execution mode, processing module 14 can specifically be used for the image block of the estimated motion vector indication that the optimum Match template is corresponding as the prediction piece, and current block is encoded or decoded.As another kind of execution mode, processing module 14 can also realize other codings or decoding function, and the below is described respectively.
The structural representation of the video image blocks processing unit that Figure 12 provides for the embodiment of the invention 11.As shown in figure 12, present embodiment is specially a code device, on the basis of above-described embodiment seven, processing module 14 further comprises: coding unit 51, be used for when estimated motion vector corresponding to optimum Match template is the first estimated motion vector, the employing decoding end motion vector DMVD technology of can deriving is encoded to current block; When estimated motion vector corresponding to optimum Match template is the second estimated motion vector, movable information is write code stream, adopt non-DMVD technology that current block is encoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector.
Processing module also can comprise this coding unit in above-described embodiment eight or nine or ten.
The structural representation of the video image blocks processing unit that Figure 13 provides for the embodiment of the invention 12.As shown in figure 13, present embodiment is specially a decoding device, and on the basis of above-described embodiment seven, processing module 14 further comprises: decoding unit 61, be used for when estimated motion vector corresponding to optimum Match template is the first estimated motion vector, adopt the DMVD technology that current block is decoded; When estimated motion vector corresponding to optimum Match template is the second estimated motion vector, according to the movable information in the code stream, adopt non-DMVD technology that current block is decoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector.
Processing module also can comprise this decoding unit in above-described embodiment eight or nine or ten.
When estimated motion vector corresponding to optimum Match template was believable estimated motion vector, above-described embodiment adopted the DMVD technology of the method for fast searching of predictability to encode or decode, and has weakened the jitter block effect.When estimated motion vector corresponding to optimum Match template is incredible estimated motion vector, if adopting in this case the DMVD technology encodes or decodes, some shakes can occur again, above-described embodiment coding side in this case, adopt non-DMVD technology to encode, decoding end is according to the movable information in the code stream (such as motion vector difference), adopt non-DMVD technology to decode, since in the non-DMVD technology according to the RDO criterion, in the destructuring zone, coding side more is ready to select less motion vector difference to encode, and can produce so the motion vector field of rule, guarantee the motion vector field regularization, also weakened the jitter block effect.
One of ordinary skill in the art will appreciate that: all or part of step that realizes said method embodiment can be finished by the relevant hardware of program command, aforesaid program can be stored in the computer read/write memory medium, this program is when carrying out, execution comprises the step of said method embodiment, and aforesaid storage medium comprises: the various media that can be program code stored such as ROM, RAM, magnetic disc or CD.
It should be noted that at last: above embodiment only in order to the technical scheme of the embodiment of the invention to be described, is not intended to limit; Although with reference to previous embodiment the embodiment of the invention is had been described in detail, those of ordinary skill in the art is to be understood that: it still can be made amendment to the technical scheme that aforementioned each embodiment puts down in writing, and perhaps part technical characterictic wherein is equal to replacement; And these modifications or replacement do not make the essence of appropriate technical solution break away from the spirit and scope of each embodiment technical scheme of the embodiment of the invention.
Claims (14)
1. a video image block processing method is characterized in that, comprising:
Obtain the matching error when the front template reference template corresponding with above estimated motion vector of current block;
Confidence level according to described estimated motion vector, the reference template that partly or entirely estimated motion vector is corresponding in the described above estimated motion vector is revised with described matching error when front template, and the confidence level of described estimated motion vector refers to estimate the consistency that the motion vector that obtains and object move really;
According to revised matching error, from an above reference template, select the optimum Match template;
According to described optimum Match template, obtain the prediction piece of current block, described current block is encoded or decoded.
2. video image block processing method according to claim 1, it is characterized in that, described confidence level according to described estimated motion vector comprises reference template and the described matching error correction of working as front template that the part estimated motion vector is corresponding in the described above estimated motion vector:
The reference template that described part estimated motion vector is corresponding and described when the default cost function of the first matching error substitution of front template, obtain the first cost function value as described revised matching error, described part estimated motion vector with a low credibility in other estimated motion vectors except described part estimated motion vector, described the first cost function value is greater than described the first matching error;
Perhaps, the reference template that described part estimated motion vector is corresponding and described when the default cost function of the second matching error substitution of front template, obtain the second cost function value as described revised matching error, described part estimated motion vector with a high credibility in other estimated motion vectors except described part estimated motion vector, described the second cost function value is less than described the second matching error;
Wherein said default cost function comprises the product of punishment parameter and described matching error.
3. video image block processing method according to claim 1, it is characterized in that, described confidence level according to described estimated motion vector comprises reference template and the described matching error correction of working as front template that all estimated motion vector is corresponding in the described above estimated motion vector:
The reference template that a described above estimated motion vector is corresponding and described when the default cost function of the matching error substitution of front template obtains cost function value as described revised matching error;
Wherein said default cost function comprises the product of punishment parameter and described matching error.
4. video image block processing method according to claim 3 is characterized in that, the value of described punishment parameter is according to the confidence level of described estimated motion vector and definite.
5. video image block processing method according to claim 4, it is characterized in that, whether the confidence level of described estimated motion vector is candidate motion vector and definite according to described estimated motion vector, described candidate motion vector with a high credibility in the confidence level of the motion vector that becomes more meticulous, the value of the punishment parameter that described candidate motion vector is corresponding is less than the value of punishment parameter corresponding to the described motion vector that becomes more meticulous;
Wherein, described candidate motion vector is known motion vector or the whole pixel precision motion vector of current block periphery; The described motion vector that becomes more meticulous is motion vector or the half-pixel accuracy motion vector that described candidate motion vector obtains by skew half-pixel accuracy position.
6. video image block processing method according to claim 4 is characterized in that, the confidence level of described estimated motion vector is definite with the similarity degree of the variance of working as front template or gradient information according to reference template corresponding to described estimated motion vector.
7. video image block processing method according to claim 4 is characterized in that, comprises described punishment parameter in the code stream.
8. video image block processing method according to claim 1 is characterized in that, and is described according to described optimum Match template, obtains the prediction piece of current block, and described current block is encoded to be comprised:
When estimated motion vector corresponding to described optimum Match template is the first estimated motion vector, adopt the decoding end motion vector DMVD technology of can deriving that described current block is encoded; When estimated motion vector corresponding to described optimum Match template is the second estimated motion vector, movable information is write code stream, adopt non-DMVD technology that described current block is encoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector.
9. video image block processing method according to claim 1 is characterized in that, and is described according to described optimum Match template, obtains the prediction piece of current block, and described current block is decoded to be comprised:
When estimated motion vector corresponding to described optimum Match template is the first estimated motion vector, adopt the DMVD technology that described current block is decoded; When estimated motion vector corresponding to described optimum Match template is the second estimated motion vector, according to the movable information in the code stream, adopt non-DMVD technology that described current block is decoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector.
10. a video image blocks processing unit is characterized in that, comprising:
Matching module is used for obtaining the matching error when the front template reference template corresponding with above estimated motion vector of current block;
Correcting module, be used for the confidence level according to described estimated motion vector, the reference template that partly or entirely estimated motion vector is corresponding in the described above estimated motion vector is revised with described matching error when front template, and the confidence level of described estimated motion vector refers to estimate the consistency that the motion vector that obtains and object move really;
Select module, be used for according to revised matching error, from an above reference template, select the optimum Match template;
Processing module is used for according to described optimum Match template, obtains the prediction piece of current block, and described current block is encoded or decoded.
11. video image blocks processing unit according to claim 10 is characterized in that, described correcting module comprises:
The first amending unit is used for the reference template that described part estimated motion vector is corresponding and described when the default cost function of the first matching error substitution of front template, obtains the first cost function value as described revised matching error; Described part estimated motion vector with a low credibility in other estimated motion vectors except described part estimated motion vector, described the first cost function value is greater than described the first matching error;
Perhaps, the second amending unit, be used for the reference template that described part estimated motion vector is corresponding and described when the default cost function of the second matching error substitution of front template, obtain the second cost function value as described revised matching error, described part estimated motion vector with a high credibility in other estimated motion vectors except described part estimated motion vector, described the second cost function value is less than described the second matching error;
Wherein said default cost function comprises the product of punishment parameter and described matching error.
12. video image blocks processing unit according to claim 10 is characterized in that, described correcting module comprises:
The 3rd amending unit is used for the reference template that a described above estimated motion vector is corresponding and described when the default cost function of the matching error substitution of front template, obtains cost function value as described revised matching error;
Wherein said default cost function comprises the product of punishment parameter and described matching error.
13. according to claim 10 or 11 or 12 described video image blocks processing unit, it is characterized in that described processing module comprises:
Coding unit is used for when estimated motion vector corresponding to described optimum Match template is the first estimated motion vector, and the employing decoding end motion vector DMVD technology of can deriving is encoded to described current block; When estimated motion vector corresponding to described optimum Match template is the second estimated motion vector, movable information is write code stream, adopt non-DMVD technology that described current block is encoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector.
14. according to claim 10 or 11 or 12 described video image blocks processing unit, it is characterized in that described processing module comprises:
Decoding unit is used for when estimated motion vector corresponding to described optimum Match template is the first estimated motion vector, adopts the DMVD technology that described current block is decoded; When estimated motion vector corresponding to described optimum Match template is the second estimated motion vector, according to the movable information in the code stream, adopt non-DMVD technology that described current block is decoded, wherein the first estimated motion vector is with a high credibility in the second estimated motion vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010144630 CN102215386B (en) | 2010-04-09 | 2010-04-09 | Video image block processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010144630 CN102215386B (en) | 2010-04-09 | 2010-04-09 | Video image block processing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102215386A CN102215386A (en) | 2011-10-12 |
| CN102215386B true CN102215386B (en) | 2013-03-27 |
Family
ID=44746483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201010144630 Active CN102215386B (en) | 2010-04-09 | 2010-04-09 | Video image block processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102215386B (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| SI3767952T1 (en) * | 2012-01-19 | 2022-01-31 | Electronics And Telecommunications Research Institute | Apparatus for image coding/decoding |
| US9325991B2 (en) * | 2012-04-11 | 2016-04-26 | Qualcomm Incorporated | Motion vector rounding |
| WO2014168444A1 (en) * | 2013-04-11 | 2014-10-16 | 엘지전자 주식회사 | Video signal processing method and device |
| CN105338351B (en) | 2014-05-28 | 2019-11-12 | 华为技术有限公司 | Intra prediction coding and decoding, array scanning method and device based on template matching |
| US10291931B2 (en) * | 2016-10-13 | 2019-05-14 | Ati Technologies Ulc | Determining variance of a block of an image based on a motion vector for the block |
| WO2018120290A1 (en) * | 2016-12-26 | 2018-07-05 | 华为技术有限公司 | Prediction method and device based on template matching |
| EP3644608A4 (en) * | 2017-07-11 | 2020-06-10 | Huawei Technologies Co. Ltd. | Decoding method and apparatus based on template matching |
| CN111201795B (en) * | 2017-10-09 | 2022-07-26 | 华为技术有限公司 | Memory access window and padding for motion vector modification |
| US11750832B2 (en) * | 2017-11-02 | 2023-09-05 | Hfi Innovation Inc. | Method and apparatus for video coding |
| CN110662033B (en) * | 2018-06-28 | 2021-11-23 | 杭州海康威视数字技术股份有限公司 | Decoding and encoding method and device thereof |
| CN110662074B (en) * | 2018-06-28 | 2021-11-23 | 杭州海康威视数字技术股份有限公司 | Motion vector determination method and device |
| EP3841751B1 (en) | 2018-09-19 | 2024-04-17 | Huawei Technologies Co., Ltd. | Method for skipping refinement based on patch similarity in bilinear interpolation based decoder-side motion vector refinement |
| CN113424535A (en) | 2019-02-13 | 2021-09-21 | 北京字节跳动网络技术有限公司 | History update based on motion vector prediction table |
| WO2020164575A1 (en) | 2019-02-14 | 2020-08-20 | Beijing Bytedance Network Technology Co., Ltd. | Decoder side motion derivation based on processing parameters |
| CN111800631A (en) * | 2020-06-22 | 2020-10-20 | 西安万像电子科技有限公司 | Data processing method and system |
| CN111711823B (en) * | 2020-06-30 | 2022-11-15 | Oppo广东移动通信有限公司 | Motion vector processing method and apparatus, electronic device, and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1925614A (en) * | 2005-08-30 | 2007-03-07 | 深圳清华大学研究院 | Rapid movement estimating method |
| CN101557514A (en) * | 2008-04-11 | 2009-10-14 | 华为技术有限公司 | Method, device and system for inter-frame predicting encoding and decoding |
| CN101605256A (en) * | 2008-06-12 | 2009-12-16 | 华为技术有限公司 | A kind of method of coding and decoding video and device |
| CN101686393A (en) * | 2008-09-28 | 2010-03-31 | 华为技术有限公司 | Fast-motion searching method and fast-motion searching device applied to template matching |
-
2010
- 2010-04-09 CN CN 201010144630 patent/CN102215386B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1925614A (en) * | 2005-08-30 | 2007-03-07 | 深圳清华大学研究院 | Rapid movement estimating method |
| CN101557514A (en) * | 2008-04-11 | 2009-10-14 | 华为技术有限公司 | Method, device and system for inter-frame predicting encoding and decoding |
| CN101605256A (en) * | 2008-06-12 | 2009-12-16 | 华为技术有限公司 | A kind of method of coding and decoding video and device |
| CN101686393A (en) * | 2008-09-28 | 2010-03-31 | 华为技术有限公司 | Fast-motion searching method and fast-motion searching device applied to template matching |
Non-Patent Citations (2)
| Title |
|---|
| Decoder side motion vector derivation for inter frame video coding;Steffen Kamp et al.;《Image Processing,2008.ICIP2008.15th IEEE International Conference on》;20081015;正文第2部分，图2，3 * |
| Steffen Kamp et al..Decoder side motion vector derivation for inter frame video coding.《Image Processing,2008.ICIP2008.15th IEEE International Conference on》.2008, |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102215386A (en) | 2011-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102215386B (en) | Video image block processing method and device | |
| CN102685504B (en) | The decoding method of video image, code device, decoding device and system thereof | |
| CN108134939B (en) | Motion estimation method and device | |
| US10051282B2 (en) | Method for obtaining motion information with motion vector differences | |
| JP2011511541A (en) | Method, apparatus, and system for interframe predictive coding | |
| TWI458356B (en) | Motion vector predictive coding method, motion vector predictive decoding method, moving picture coding apparatus, moving picture decoding apparatus, and programs thereof | |
| TWI442776B (en) | Motion vector predictive coding method, motion vector predictive decoding method, moving picture coding apparatus, moving picture decoding apparatus, and programs thereof | |
| TW202019179A (en) | Method and apparatus of motion-vector rounding unification for video coding system | |
| JP2020150313A (en) | Image decoder, image decoding method, and program | |
| CN113794888B (en) | Decoding method, encoding method, device, equipment and machine readable storage medium | |
| CN110662033B (en) | Decoding and encoding method and device thereof | |
| MX2021000351A (en) | Video encoder, video decoder, and corresponding method. | |
| JP6019797B2 (en) | Moving picture coding apparatus, moving picture coding method, and program | |
| CN104202606A (en) | A starting point determination method based on HEVC motion estimation | |
| CN110636290B (en) | Video image motion vector encoding/decoding method and encoder/decoder | |
| CN110662074B (en) | Motion vector determination method and device | |
| KR101502145B1 (en) | Method and apparatus for video encoding and decoding | |
| CN112702601B (en) | Method and apparatus for determining motion vector for inter prediction | |
| CN110868611B (en) | Video encoding and decoding methods and devices | |
| CN110365987B (en) | Motion vector determination method, device and equipment | |
| WO2020255846A1 (en) | Image decoding device, image decoding method, and program | |
| CN110691247B (en) | Decoding and encoding method and device | |
| KR102651484B1 (en) | Method and apparatus for coding images of video sequences, and terminal device | |
| CN110868612A (en) | Video encoding method and device and decoding method and device | |
| JP2006108932A (en) | Motion picture encoding device using plural reference pictures, motion picture decoding device, motion picture encoding method, motion picture decoding method, motion picture encoding program and motion picture decoding program |
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 |