CN104244003B - A kind of method and device for determining Motion Vector Cost - Google Patents
A kind of method and device for determining Motion Vector Cost Download PDFInfo
- Publication number
- CN104244003B CN104244003B CN201410406661.4A CN201410406661A CN104244003B CN 104244003 B CN104244003 B CN 104244003B CN 201410406661 A CN201410406661 A CN 201410406661A CN 104244003 B CN104244003 B CN 104244003B
- Authority
- CN
- China
- Prior art keywords
- value
- values
- look
- cost
- benchmark
- 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
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
It is easy to the hard-wired method and device for determining Motion Vector Cost the invention discloses a kind of.The method for determining Motion Vector Cost, including：Each macro block being divided into for present frame to be encoded, the quantization parameter QP values used according to current macro, choose the look-up table needed to use, the COST values that MV each span and each span are mapped are recorded in the look-up table, and with MV value change from small to large, MV span becomes larger；The MV values that are used according to current macro and the look-up table used the need for selecting, determine the corresponding COST values of the MV values.Using this programme, the complexity of the hardware logic of Motion Vector Cost calculating section can be effectively reduced, so as to obtain relatively low power consumption and less silicon area.
Description
Technical field
The present invention relates to motion estimation techniques field, more particularly to a kind of method and device for determining Motion Vector Cost.
Background technology
Estimation (Motion Estimation) is widely used technology in video compress, while being video compress
During one of most crucial and operand the best part, it is how high-quality most important with efficient progress estimation.
The basic thought of estimation is the macro block (Macroblocks) that each frame of image sequence is divided into many non-overlapping copies, and
Think that the displacement of all pixels in macro block is all identical, it is then a certain to each macro block to reference frame (REF) given specific to search
The block most like with current block, i.e. match block, match block and the phase of current block are found out in the range of rope according to certain matching criterior
It is motion vector (Motion Vector, MV) to displacement.When video compress, only motion vector and residual error number need to be preserved
According to can just recover current block completely.
H.264 it is ITU (International Telecommunication Unite, International Telecommunications Union) and MPEG
The video encoding standard that (Motion Picture Experts Group, Motion Picture Experts Group) joint is formulated.X264 is out
The realization of the H.264 codec of source and generally acknowledged first water, currently acquired extensive accreditation is with using, even if such as
This, huge operand causes high-resolution real-time video to be recorded in software view and be also difficult to, and application specific integrated circuit is then
It is one of more feasible implementation.But how to realize that so huge logic scale also has larger choose in circuit-level
War, is combined the characteristics of by algorithm with hardware, therefrom finding cleverly method becomes one of direction that everybody makes great efforts to study.
The motion estimation algorithm based on best matching blocks is employed in X264, the process of estimation is to calculate with seeking
The process of blocks and optimal matching blocks cost (COST).Residual error data cost generally by current block of the cost of match block, motion vector generation
Valency is constituted with reference frame cost.The cost of match block represents the cost that code word is spent needed for after current block coding to a certain extent
(size after video compress is made up of code word one by one), cost is more low, means the current block under current motion vector
Better encoding efficiency can be obtained.The process for calculating each cost is respectively provided with certain complexity.
Motion Vector Cost is determined by formula [1] in X264：
MV_COST=lambda* (log2f (MV |+1) * 2+0.718+！！i)+5f [1]
Wherein, MV_COST represents Motion Vector Cost, and lambda represents Lagrange coefficient, by the amount in cataloged procedure
Change parameter (Quantization Parameter, QP) mapping to obtain.From above-mentioned formula [1] it can be seen that being included in calculating process
There is Logarithmic calculation, Logarithmic calculation operand is very huge, therefore, the initial phase in X264 further in coding in advance will
COST values corresponding with QP all MV are precalculated and preserved in a lookup table, are led in follow-up actual motion estimation process
Cross to table look-up and directly obtain.Wherein, one look-up table of each QP value correspondence, each look-up table is used to represent between MV and COST
Corresponding relation, be in the prior art, between MV and COST one-to-one relationship.
This processing mode is no doubt a good method in software view, but this will set to hardware in hardware view
Meter brings certain challenge, it is contemplated that H.264 QP span is 0~51 in standard, accordingly, 52 look-up tables of correspondence；
And MV span is bigger, by taking MV ± 2048 as an example, then the size of whole look-up table will exceed 200Kbyte.Specific implementation
In, if can be due to Logarithmic calculation so that hardware circuit becomes not sounding feasible not by the way of look-up table uses dynamic calculation
Border；According to look-up table, the hardware resource of bulk is taken because of the size of table if look-up table is placed in inside hardware, such as will
Look-up table, which is placed in external memory dynamic value, will then consume extra external memory bandwidth, and again can because of external memory access delay influence property
Energy.To sum up, it would be highly desirable to provide it is a kind of be easy to it is hard-wired determination Motion Vector Cost scheme.
The content of the invention
Offer of the embodiment of the present invention is a kind of to be easy to the hard-wired method and device for determining Motion Vector Cost, to have
The complexity of the hardware logic of effect reduction Motion Vector Cost calculating section, so as to obtain relatively low power consumption and less silicon
Piece area.
The embodiment of the present invention provides a kind of method for determining Motion Vector Cost, including：
Each macro block being divided into for present frame to be encoded, the quantization parameter QP values used according to current macro,
Record MV each span and each span is mapped in the look-up table that selection is needed to use, the look-up table
COST values, and with MV value change from small to large, MV span becomes larger；
The MV values that are used according to current macro and the look-up table used the need for selecting, determine the MV values correspondence
COST values.
Based on same technical concept, the embodiment of the present invention provides a kind of device for determining Motion Vector Cost, including：
Memory module, respectively taking for MV is recorded for storing in the corresponding look-up table of each quantization parameter QP values, the look-up table
The COST values that value scope and each span are mapped, and with MV value change from small to large, MV value
Scope becomes larger；
Module is chosen, for each macro block being divided into for present frame to be encoded, is used according to current macro
Quantization parameter QP values, choose the look-up table needed to use from the memory module；
Determining module, the look-up table used for the MV values used according to current macro and the need for selecting, it is determined that
The corresponding COST values of the MV values.
The method and device provided in an embodiment of the present invention for determining Motion Vector Cost, using the look-up table of special construction,
Record MV each span and the COST values that are mapped of each span in look-up table, and as MV value is from small
To big change, MV span becomes larger, will MV values and the one-to-one mapping relations of COST values in the prior art, change
It is changed into fixing some MV in span mapping relations corresponding with a COST value, and the larger MV of value is corresponding
MV span is also larger, so as to keep effective computational accuracy significantly reducing while individual searches table size.Due to
Individual searches significantly reducing for table size, so as to effectively reduce the size of whole look-up table so that Motion Vector Cost is calculated
The complexity reduction of partial hardware logic, so as to obtain relatively low power consumption and less silicon area.
Other features and advantage will be illustrated in the following description, also, partly be become from specification
Obtain it is clear that or being understood by implementing the application.The purpose of the application and other advantages can be by the explanations write
Specifically noted structure is realized and obtained in book, claims and accompanying drawing.
Brief description of the drawings
Accompanying drawing is used for providing a further understanding of the present invention, and constitutes a part for specification, with present invention implementation
Example is used to explain the present invention together, is not construed as limiting the invention.In the accompanying drawings：
Fig. 1 is the method flow diagram of determination Motion Vector Cost in the embodiment of the present invention；
Fig. 2 is the device block diagram of determination Motion Vector Cost in the embodiment of the present invention；
Fig. 3 is a kind of possible structural representation of acquisition module in the embodiment of the present invention.
Embodiment
Offer of the embodiment of the present invention is a kind of to be easy to the hard-wired method and device for determining Motion Vector Cost, to have
The complexity of the hardware logic of effect reduction Motion Vector Cost calculating section, so as to obtain relatively low power consumption and less silicon
Piece area.The preferred embodiments of the present invention are illustrated below in conjunction with Figure of description, it will be appreciated that described herein excellent
Select embodiment to be merely to illustrate and explain the present invention, be not intended to limit the present invention.And in the case where not conflicting, the application
In embodiment and the feature in embodiment can be mutually combined.
The scheme provided in an embodiment of the present invention for determining Motion Vector Cost, will be to Motion Vector Cost meter in estimation
Calculate part to carry out simplifying design, specifically include：
Simplify layout strategy one
The QP (quantization parameter) of the different numerical value used in view of a frame in actual coding number generally will not be too many,
In the embodiment of the present invention, setting only can be in benchmark in the benchmark QP of a frame, and configuration baseline QP fluctuation range, cataloged procedure
On the basis of QP, fluctuation is using the QP values of limited quantity up and down in fluctuation range, each QP in benchmark QP and its fluctuation range
Value corresponds to a look-up table respectively, by controlling fluctuation range, can control QP quantity, the QP of limited quantity causes look-up table
Quantity reduce, so as to effectively reduce the size of whole look-up table.
Simplify layout strategy two
In view of the actual distribution rule of MV (motion vector) in common actual video recording process, usual MV values compared with
The proportion that small number is accounted for is higher, in addition because the introducing of logarithm, the change of COST values have during calculating COST (cost)
The characteristics of increasingly being eased up with MV value change from small to large.Based on above-mentioned consideration, the embodiments of the invention provide
The COST that MV each span and each span are mapped is recorded in the look-up table of a set of special construction, look-up table
Value, and with MV value change from small to large, MV span becomes larger.To MV values in the prior art and COST values
One-to-one mapping relations, mapping relations corresponding with a COST value some MV for changing into fixed span,
And the corresponding MV of the larger MV of value span is also larger, so as to be protected significantly reducing while individual searches table size
Hold effective computational accuracy.
In specific implementation, look-up table can use a kind of preferred construction, i.e., with MV value change from small to large, MV
Span be in 2 exponential function trend growth, as shown in table 1, illustrated by taking MV values 0~2047 as an example.
Table 1
Above-mentioned table 1 illustrates the lookup rule of the COST values corresponding to MV, and each cell represents depositing for COST value
Storage, the numerical value of the inside represents to be mapped to the MV of this COST values span, it can be seen that：
When MV value is 0~31, MV span corresponds for 1, i.e. MV values with COST values；
When MV value is 32~63, MV span is 2, i.e. each two MV one COST value of correspondence；
When MV value is 64~127, MV span is 4, i.e. one COST value of every four MV correspondences；
When MV value is 128~255, MV span is 8, i.e. one COST value of every eight MV correspondences；
When MV value is 256~511, MV span is 16, i.e. one COST value of every 16 MV correspondences；
When MV value is 512~1023, MV span is 32, i.e. one COST of every 32 MV correspondences
Value；
When MV value is 1024~2047, MV span is 64, i.e. one COST of every 64 MV correspondences
Value；
The rest may be inferred, and precision is gradually reduced.
The look-up table of special construction provided in an embodiment of the present invention, is searched correspondingly compared to all MV values and COST values
Table, the size required for individual look-up table will be significantly reduced, for example the look-up table shown in table 1, wherein each grid represents 1
16bit memory space, then look-up table only need 128 16bit memory space；Simultaneously in view of the distribution character of COST values,
The precision influence brought in larger MV is also relatively limited.
Certainly, table 1 is only to be directed to a kind of preferred construction of look-up table for example, in specific implementation, look-up table can not
It is limited to the concrete structure of table 1, if what each span and each span that MV is recorded in look-up table were mapped
COST values, and as MV value change from small to large, MV span become larger, can reach reduction look-up table
Size and the purpose for keeping computational accuracy.
In specific implementation, above-mentioned simplified layout strategy one and simplified layout strategy two can be used cooperatively, can also be independent
Use.Used cooperatively preferably, simplifying layout strategy one and simplifying layout strategy two, for example, based on Finite Number described above
The simplification layout strategy one of the QP values of amount, using benchmark QP fluctuation range as [- 4,3] exemplified by, it is assumed that benchmark QP values be 30, then
Each QP value is respectively 26,27,28,29,30,31,32,33 in benchmark QP and its fluctuation range, then accordingly, it is only necessary to 8
Look-up table；Further, if every look-up table, which is used in the special construction described in table 1, hardware, only needs to 8*128*2=
2Kbyte (8bit=1byte) whole lookup table size, and will exceed according to the size of the whole look-up table of prior art
200Kbyte, it is seen that greatly reduce lookup table size.
Based on the introduction of above-mentioned simplified layout strategy, the embodiments of the invention provide a kind of side for determining Motion Vector Cost
Method, as shown in figure 1, including：
S101, each macro block being divided into for present frame to be encoded, the QP values used according to current macro are chosen
The look-up table needed to use, wherein, record MV each span and each span is mapped in look-up table
COST values, and with MV value change from small to large, MV span becomes larger.
Preferably, with MV value change from small to large, MV span can increase in 2 exponential function trend
It is long, wherein：
When MV value is 0~31, each MV one COST value of correspondence；
When MV value is 32~63, each two MV one COST value of correspondence；
When MV value is 64~127, one COST value of every four MV correspondences；
When MV value is 128~255, one COST value of every eight MV correspondences；
When MV value is 256~511, one COST value of every 16 MV correspondences；
When MV value is 512~1023, one COST value of every 32 MV correspondences；
When MV value is 1024~2047, one COST value of every 64 MV correspondences；
The rest may be inferred.
, can be with unanimously, prestoring each in QP span in the prior art in S101 specific implementation
The corresponding look-up table of QP values, for example H.264 QP span is 0~51 in standard, accordingly, prestores each QP value
Corresponding look-up table, that is, store 52 look-up tables；On this basis, the QP values used according to current macro, selection is needed to use
Look-up table, specifically include：The QP values used according to current macro, choose the corresponding lookup of QP values that the current macro is used
Table.Choosing method with unanimously, specifically repeating no more in the prior art.
In S101 specific implementation, simplifying layout strategy one and simplifying layout strategy two to use cooperatively, Ke Yi
Benchmark QP and its fluctuation range are pre-set in hardware by the way of software merit rating, then hardware makes according to current macro is actual
QP values and benchmark QP difference are indexed, and which table selection specifically needs to use (with benchmark QP and its fluctuation range
Exemplified by 8 QP values, then 8 look-up tables are corresponded to), the lookup corresponding to the QP of difference index can be chosen in specific implementation
Table, then further finds current COST values further according to the MV values that current macro is used.
It should be noted that the fluctuation range for assuming benchmark QP is [- 4,3], then need to limit in advance each in present frame
For the QP relative datums QP that individual macro block is used only can in the range of [benchmark QP-4, benchmark QP+3] value, so only with 8
Look-up table, the QP values used using current macro and one in benchmark QP difference demapping this 8 tables.
Preferably, benchmark QP fluctuation range is defined to [- 4,3] or [- 3,4].Compared to QP in existing H.264 standard
Span be 0~51, accordingly, correspondence 52 look-up tables；During the present invention is implemented, QP preferred span is benchmark
QP-4~benchmark QP+3 or benchmark QP-3~benchmark QP+4, accordingly, 8 look-up tables of correspondence greatly reduce look-up table
Quantity, so as to effectively reduce the size of whole look-up table.
S102, the MV values used according to current macro and the corresponding look-up table of QP values selected, it is determined that current grand
The corresponding COST values of MV values that block is used.
In specific implementation, because look-up table is used to represent the corresponding relation between MV and COST, so according to current macro
In the MV values used, the look-up table that can be used the need for selecting, determine that the MV values that current macro is used are corresponding
COST values, specific search procedure with unanimously, repeating no more in the prior art.
Based on same technical concept, the embodiments of the invention provide a kind of device for determining Motion Vector Cost, due to this
The principle that device solves problem is consistent with the method for determining Motion Vector Cost, therefore the implementation of the device may refer to method
Implementation, repeat part do not repeating.
As shown in Fig. 2 the device provided in an embodiment of the present invention for determining Motion Vector Cost, including：
Memory module 201, for storing the corresponding look-up table of each QP values, wherein, MV each value model is recorded in look-up table
Enclose and COST values that each span is mapped, and with MV value change from small to large, MV span
Become larger；
Module 202 is chosen, for each macro block being divided into for present frame to be encoded, is used according to current macro
QP values, the look-up table needed to use is chosen from memory module 201；
Determining module 203, the look-up table used for the MV values used according to current macro and the need for selecting,
Determine the corresponding COST values of MV values that current macro is used.
Preferably, in the look-up table that memory module 201 is stored, with MV value change from small to large, MV value
Scope is in 2 exponential function trend growth, wherein：
When MV value is 0~31, each MV one COST value of correspondence；
When MV value is 32~63, each two MV one COST value of correspondence；
When MV value is 64~127, one COST value of every four MV correspondences；
When MV value is 128~255, one COST value of every eight MV correspondences；
When MV value is 256~511, one COST value of every 16 MV correspondences；
When MV value is 512~1023, one COST value of every 32 MV correspondences；
When MV value is 1024~2047, one COST value of every 64 MV correspondences；
The rest may be inferred.
Store each in the benchmark QP configured in advance for present frame and its fluctuation range in specific implementation, in memory module 201
The corresponding look-up table of QP values；And
Module 202 is chosen, specifically for the QP values used according to current macro and the benchmark configured in advance for present frame
QP and its fluctuation range, choose the look-up table needed to use from memory module 201.
Wherein, a kind of possible structure of module 202 is chosen, as shown in figure 3, can specifically include：
Computing unit 221, for calculating the difference between the QP values and benchmark QP that current macro uses, wherein, present frame
In the QP that uses of each macro block values in benchmark QP and its fluctuation range limited range；
Indexing units 222, the difference for being calculated according to computing unit 221 is carried out in benchmark QP and its fluctuation range
Index, chooses the look-up table corresponding to the QP of difference index as the look-up table needed to use from memory module 201.
Preferably, benchmark QP fluctuation range includes [- 4,3] or [- 3,4].Compared to QP in existing H.264 standard
Span is 0~51, accordingly, 52 look-up tables of correspondence；During the present invention is implemented, QP preferred span is benchmark QP-4
~benchmark QP+3 or benchmark QP-3~benchmark QP+4, accordingly, 8 look-up tables of correspondence greatly reduce the quantity of look-up table,
So as to effectively reduce the size of whole look-up table.
During those skilled in the art are it should be appreciated that above-mentioned Module Division mode is only numerous Module Division modes
One kind, if being divided into other modules or non-division module, has above-mentioned functions, all as long as determining the device of Motion Vector Cost
Should be within the protection domain of the application.
The method and device provided in an embodiment of the present invention for determining Motion Vector Cost, using the look-up table of special construction,
Record MV each span and the COST values that are mapped of each span in look-up table, and as MV value is from small
To big change, MV span becomes larger, will MV values and the one-to-one mapping relations of COST values in the prior art, change
It is changed into fixing some MV in span mapping relations corresponding with a COST value, and the larger MV of value is corresponding
MV span is also larger, so as to keep effective computational accuracy significantly reducing while individual searches table size.Due to
Individual searches significantly reducing for table size, so as to effectively reduce the size of whole look-up table so that Motion Vector Cost is calculated
The complexity reduction of partial hardware logic, so as to obtain relatively low power consumption and less silicon area.
Further, the method and device provided in an embodiment of the present invention for determining Motion Vector Cost, it is contemplated that actual to compile
The QP for the different numerical value that a frame in is used number generally will not be too many in code, therefore presets the benchmark QP of a frame, encodes
During only can on the basis of benchmark QP, in fluctuation range up and down fluctuation using limited quantity QP values, due to benchmark QP
And its each QP values in fluctuation range correspond to a look-up table respectively, the QP values of limited quantity cause the quantity of look-up table to reduce,
Further effectively reduce the size of whole look-up table so that the complexity of the hardware logic of Motion Vector Cost calculating section is entered
One step is reduced, so as to obtain relatively low power consumption and less silicon area.
, but those skilled in the art once know basic creation although preferred embodiments of the present invention have been described
Property concept, then can make other change and modification to these embodiments.So, appended claims are intended to be construed to include excellent
Select embodiment and fall into having altered and changing for the scope of the invention.
Obviously, those skilled in the art can carry out the essence of various changes and modification without departing from the present invention to the present invention
God and scope.So, if these modifications and variations of the present invention belong to the scope of the claims in the present invention and equivalent technologies
Within, then the present invention is also intended to comprising including these changes and modification.
Claims (8)
1. a kind of method for determining Motion Vector Cost, it is characterised in that including：
Each macro block being divided into for present frame to be encoded, the quantization parameter QP values used according to current macro are chosen
The COST that MV each span and each span are mapped is recorded in the look-up table needed to use, the look-up table
Value, and with MV value change from small to large, MV span becomes larger；
The MV values that are used according to current macro and the look-up table used the need for selecting, determine that the MV values are corresponding
COST values；
With MV value change from small to large, MV span is in 2 exponential function trend growth, wherein：
When MV value is 0~31, each MV one COST value of correspondence；
When MV value is 32~63, each two MV one COST value of correspondence；
When MV value is 64~127, one COST value of every four MV correspondences；
When MV value is 128~255, one COST value of every eight MV correspondences；
When MV value is 256~511, one COST value of every 16 MV correspondences；
When MV value is 512~1023, one COST value of every 32 MV correspondences；
When MV value is 1024~2047, one COST value of every 64 MV correspondences；
The rest may be inferred.
2. the method as described in claim 1, it is characterised in that be in advance present frame configuration baseline QP and its fluctuation range, and
And the corresponding look-up table of each QP values in only Memory Reference QP and its fluctuation range, and
The quantization parameter QP values used according to current macro, choose the look-up table needed to use, specifically include：
The quantization parameter QP values used according to current macro and the benchmark QP and its fluctuation range that are configured in advance for present frame,
The look-up table needed to use is chosen from the corresponding look-up table of each QP values in the benchmark QP and its fluctuation range.
3. method as claimed in claim 2, it is characterised in that the QP values used according to current macro and be in advance
The benchmark QP and its fluctuation range of present frame configuration, from the corresponding look-up table of each QP values in the benchmark QP and its fluctuation range
The look-up table needed to use is chosen, is specifically included：
The difference between the QP values that use of current macro and the benchmark QP is calculated, wherein, the macro block of each in present frame is used
QP in the benchmark QP and its fluctuation range limited range value；
It is indexed according to the difference in the benchmark QP and its fluctuation range, corresponding to the QP for choosing the difference index
Look-up table be used as the look-up table needed to use.
4. method as claimed in claim 3, it is characterised in that the fluctuation range of the benchmark QP include [- 4,3] or [- 3,
4]。
5. a kind of device for determining Motion Vector Cost, it is characterised in that including：
Memory module, MV each value model is recorded for storing in the corresponding look-up table of each quantization parameter QP values, the look-up table
Enclose and COST values that each span is mapped, and with MV value change from small to large, MV span
Become larger；
Choose module, for each macro block being divided into for present frame to be encoded, the quantization used according to current macro
Parameter QP values, choose the look-up table needed to use from the memory module；
Determining module, the look-up table used for the MV values used according to current macro and the need for selecting, it is determined that described
The corresponding COST values of MV values；
In the look-up table of memory module storage, with MV value change from small to large, MV span is in 2 finger
Number function trend growth, wherein：
When MV value is 0~31, each MV one COST value of correspondence；
When MV value is 32~63, each two MV one COST value of correspondence；
When MV value is 64~127, one COST value of every four MV correspondences；
When MV value is 128~255, one COST value of every eight MV correspondences；
When MV value is 256~511, one COST value of every 16 MV correspondences；
When MV value is 512~1023, one COST value of every 32 MV correspondences；
When MV value is 1024~2047, one COST value of every 64 MV correspondences；
The rest may be inferred.
6. device as claimed in claim 5, it is characterised in that
Store and searched in advance for each QP values are corresponding in the benchmark QP and its fluctuation range of present frame configuration in the memory module
Table；And
The selection module, is specifically for the quantization parameter QP values that are used according to current macro and in advance present frame configuration
Benchmark QP and its fluctuation range, the look-up table needed to use is chosen from the memory module.
7. device as claimed in claim 6, it is characterised in that the selection module, is specifically included：
Computing unit, for calculating the difference between the QP values and the benchmark QP that current macro uses, wherein, it is every in present frame
The QP that one macro block is used values in the benchmark QP and its fluctuation range limited range；
Indexing units, the difference for being calculated according to the computing unit carries out rope in the benchmark QP and its fluctuation range
Draw, the look-up table corresponding to the QP of the difference index is chosen from the memory module as the look-up table needed to use.
8. device as claimed in claim 7, it is characterised in that the fluctuation range of the benchmark QP include [- 4,3] or [- 3,
4]。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410406661.4A CN104244003B (en) | 2014-08-18 | 2014-08-18 | A kind of method and device for determining Motion Vector Cost |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410406661.4A CN104244003B (en) | 2014-08-18 | 2014-08-18 | A kind of method and device for determining Motion Vector Cost |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104244003A CN104244003A (en) | 2014-12-24 |
| CN104244003B true CN104244003B (en) | 2017-08-15 |
Family
ID=52231219
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410406661.4A Active CN104244003B (en) | 2014-08-18 | 2014-08-18 | A kind of method and device for determining Motion Vector Cost |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104244003B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104185029B (en) * | 2014-08-18 | 2017-09-29 | 北京君正集成电路股份有限公司 | The method and device of Motion Vector Cost is determined in a kind of estimation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101175210A (en) * | 2006-10-30 | 2008-05-07 | 中国科学院计算技术研究所 | Entropy decoding method and device used for decoding video estimation residual error coefficient |
| CN101754020A (en) * | 2008-12-15 | 2010-06-23 | 辉达公司 | Method and system for encoded video compression |
| CN102263947A (en) * | 2010-05-27 | 2011-11-30 | 香港科技大学 | Method and system for motion estimation of images |
| CN102413323A (en) * | 2010-01-13 | 2012-04-11 | 中国移动通信集团广东有限公司中山分公司 | H.264-based video compression method |
| CN103636202A (en) * | 2011-06-22 | 2014-03-12 | 高通股份有限公司 | Quantization parameter prediction in video coding |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7362809B2 (en) * | 2003-12-10 | 2008-04-22 | Lsi Logic Corporation | Computational reduction in motion estimation based on lower bound of cost function |
| JP5664762B2 (en) * | 2011-03-17 | 2015-02-04 | 富士通株式会社 | Moving picture decoding method, moving picture encoding method, moving picture decoding apparatus, and moving picture decoding program |
-
2014
- 2014-08-18 CN CN201410406661.4A patent/CN104244003B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101175210A (en) * | 2006-10-30 | 2008-05-07 | 中国科学院计算技术研究所 | Entropy decoding method and device used for decoding video estimation residual error coefficient |
| CN101754020A (en) * | 2008-12-15 | 2010-06-23 | 辉达公司 | Method and system for encoded video compression |
| CN102413323A (en) * | 2010-01-13 | 2012-04-11 | 中国移动通信集团广东有限公司中山分公司 | H.264-based video compression method |
| CN102263947A (en) * | 2010-05-27 | 2011-11-30 | 香港科技大学 | Method and system for motion estimation of images |
| CN103636202A (en) * | 2011-06-22 | 2014-03-12 | 高通股份有限公司 | Quantization parameter prediction in video coding |
Non-Patent Citations (1)
| Title |
|---|
| 分数像素运动矢量代价产生器的VLSI设计;王庆春;《电视技术 数字电视与数字视频_实用设计》;20071231;第31卷(第04期);第1节第1-3段，第2节第1-3段，第3节第1段，第4节第1段，图2 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104244003A (en) | 2014-12-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6841361B2 (en) | Logical intramode naming in HEVC video coding | |
| Fuchs | Spread representations | |
| JP2003204552A (en) | Improved block transform and quantization for image and video coding | |
| CN105554510A (en) | Method and apparatus for decoding video | |
| US20140010284A1 (en) | Image transform and inverse transform method, and image encoding and decoding device using same | |
| KR102218196B1 (en) | Encoder, method thereof, and apparatuses including the encoder | |
| CN105340274A (en) | Depth map encoding method and apparatus thereof, and depth map decoding method and an apparatus thereof | |
| KR102365685B1 (en) | Method for operating of encoder, and devices having the encoder | |
| CN105993174A (en) | Video encoding method and apparatus, and video decoding method and apparatus, for signaling sao parameter | |
| CN104244003B (en) | A kind of method and device for determining Motion Vector Cost | |
| CN108322749A (en) | The coefficient optimization method of RDOQ, the accelerating method and device of RDOQ | |
| Bhatia et al. | AMM: Adaptive multilinear meshes | |
| Piyasena et al. | Reducing dynamic power in streaming CNN hardware accelerators by exploiting computational redundancies | |
| CN104185029B (en) | The method and device of Motion Vector Cost is determined in a kind of estimation | |
| CN116389752A (en) | Video encoding method and device, video decoding method and device and electronic equipment | |
| KR100968373B1 (en) | Method of clustering variable length code table, and method and apparatus for sharing memory in multi codec using the same | |
| US20150170754A1 (en) | Mitigating Inter-Cell Interference | |
| JP6412589B2 (en) | Apparatus, computer program, and computer-implemented method | |
| CN113887719B (en) | Model compression method and device | |
| CN105684315B (en) | Method and apparatus for vector segmentation for coding | |
| TW201038081A (en) | Circuit and method for multi-format video codec | |
| KR20160011782A (en) | Video encoding circuit and video encoding method therewith | |
| CN116339627A (en) | Data fuzzy processing method and device suitable for low-precision storage equipment | |
| CN115660070A (en) | Compression method for confrontation generation neural network | |
| Kang et al. | Analysis and solution of CNN accuracy reduction over channel loop tiling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |