CN103581504B - Motion compensation image processor and image treatment method - Google Patents
Motion compensation image processor and image treatment method Download PDFInfo
- Publication number
- CN103581504B CN103581504B CN201210251213.2A CN201210251213A CN103581504B CN 103581504 B CN103581504 B CN 103581504B CN 201210251213 A CN201210251213 A CN 201210251213A CN 103581504 B CN103581504 B CN 103581504B
- Authority
- CN
- China
- Prior art keywords
- motion
- vector
- image
- image block
- module
- 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.)
- Expired - Fee Related
Links
Abstract
The present invention relates to motion compensation image processor, comprising external memory storage, memory cache, motion compensation module, judge module and acquisition module.External memory storage is used to store the reference picture related to an image block.Motion compensation module moves compensation to image block before being sequentially directed to one and the image block.When the motion compensation module moves compensation program to the preceding image block, judge module judges motion-vector of the image block relative to reference picture.Before the motion compensation module moves compensation to the image block, the acquisition module captures to memory cache the reference zone in reference picture corresponding to the motion-vector from external memory storage.
Description
Technical field
The present invention is related to image processing technique, and especially with manage/use image processing system in memory
Technology is related.
Background technology
Motion compensation (motion compensation) is a kind of machine for being widely used in dynamic video compression field
System.The block type movement benefit that (moving picture experts group, MPEG) specification is used is knitted with animation expert group
As a example by repaying, picture to be encoded is divided into the block that multiple sizes are the pixels of 16 pixel * 16；For each block, encoder
A most like reference zone can be found out from reference picture, and judge the shifting between a block and its corresponding reference zone
Moving vector.In addition to motion-vector, encoder can also judge that the presentation content between a block and its corresponding reference zone is poor
It is different, referred to as amount of redundancy (residual).After one block deducts the factor of motion-vector, amount of redundancy more similar to its reference zone
It is smaller.In coding result, each block is represented with its motion-vector and amount of redundancy.
Fig. 1 (A) is used to represent work schedule example of the present code device when compensation is moved.Encoder first when
Between the most like reference zone of block A in picture to be encoded is judged in section T1.Due to being compared with picture to be encoded
Reference picture is normally stored in the external memory storage of encoder, it is assumed that time section T1 inner encoders judge with it is to be encoded
Block A in picture most like reference zone is distributed in block R1 and block R2 in reference picture.Correspond to area producing
During the amount of redundancy of block A, encoder must capture the image data of block R1 and block R2 from external memory storage.Time section T2 generations
Table encoder sends out the time for waiting external memory storage to respond after the requirement for capturing block R1.Time section T3 is represented and deposited from outside
Block R1 is transferred to reservoir the time of encoder.Time section T4 is represented after encoder sends out the requirement for capturing block R2 and waited
The time that external memory storage is responded.Time section T5 represents the time that block R2 is transferred to encoder from external memory storage.When
Between section T6 be then to represent the time that encoder produces the amount of redundancy corresponding to block A according to block R1 and block R2.Class according to this
Push away, when encoder will be each block generation amount of redundancy in picture to be encoded, it is necessary to consume six times such as T1 ~ T6.
Fig. 1 (B) is used to represent work schedule example of the existing decoder when compensation is moved.Similarly, to
The reference picture that picture to be decoded is compared is normally stored in the external memory storage of decoder.When being intended to rebuild foregoing block A, solution
Code device judges to produce the reference of this motion-vector according to this first in the motion-vector in time section T1 ' in decoded result
Block R1 and block R2 of the area distribution in reference picture.Therefore, decoder must capture block R1 from its external memory storage
With the image data of block R2.Time section T2 ' is represented after decoder sends out the requirement for capturing block R1 and is waited external memory storage
The time of response.Time section T3 ' represents the time that block R1 is transferred to decoder from external memory storage.Time section T4 '
Represent decoder and send out the time for waiting external memory storage to respond after the requirement for capturing block R2.Time section T5 ' is represented outside
Block R2 is transferred to portion's memory the time of decoder.Time section T6 ' is then to represent decoder according to block R1, block R2
And amount of redundancy rebuilds the time of block A.The rest may be inferred, during each block that decoder will be rebuild in picture to be decoded, all must
Six times such as T1 ' ~ T6 ' must be consumed.
One of development trend of image system is to be continuously increased picture dimension and resolution now, and encoder and decoder are all
Therefore it is required with operation efficiency higher.Be can be seen that by described above, the total length of time section T1 ~ T6 and T1 ' ~ T6 ' is healed
Long, the operation efficiency of encoder and decoder is poorer.
The content of the invention
To solve the above problems, the present invention proposes a kind of motion compensation image processor and motion compensation image processing side
Method, by when imposing to an image block motion compensation program simultaneously for next image block in advance will from external memory storage
Institute's palpus reference data is captured to memory cache, reaches timesaving effect.Additionally, image processor of the invention and shadow
As processing method thereby can further lift treatment effect using the concept of the data for not repeating to have had in acquisition memory cache
Rate.
A specific embodiment of the invention is a kind of motion compensation image processor, wherein comprising external storage
Device, memory cache, motion compensation module, judge module and acquisition module.External memory storage is used to store and an image block
Related reference picture.Motion compensation module moves compensation to image block before being sequentially directed to one and the image block.
When the motion compensation module moves compensation to the preceding image block, judge module judges the image block relative to reference
The motion-vector of picture.Before the motion compensation module moves compensation to the image block, the acquisition module will be with reference to picture
A reference zone in face corresponding to the motion-vector is captured to memory cache from external memory storage.
Another specific embodiment of the invention be it is a kind of with unification external memory storage and a memory cache running
Motion compensation image treatment method.The external memory storage stores a reference picture related to an image block.The method is first
A motion compensation step is performed, compensation is moved to image block before.Before the motion compensation step terminates, the method is opened
Beginning performs one and judges step, judges a motion-vector of the image block relative to the reference picture.Then, the method performs one
Step is captured, the reference zone in the reference picture corresponding to the motion-vector is captured to the cache from the external memory storage
Memory.Then, the method performs another motion compensation step, and compensation is moved to the image block.
Another specific embodiment of the invention is a kind of motion compensation image processor, wherein comprising external storage
Device, memory cache, motion compensation module, judge module and acquisition module.External memory storage is used to store and an image block
A related reference picture.Motion compensation module is used to move compensation for the image block.Judge module judges the shadow
A motion-vector as block relative to the reference picture.Acquisition module will correspond to the motion-vector from the external memory storage
One reference zone and the remaining area that is not stored in the memory cache are captured to the memory cache.
Another specific embodiment of the invention be it is a kind of with unification external memory storage and a memory cache running
Motion compensation image treatment method.The external memory storage is used to store a reference picture related to an image block.The method
A judgement step is first carried out, a motion-vector of the image block relative to the reference picture is judged.Then, the method is performed
One captures step, and the reference zone for corresponding to the motion-vector is not stored in into the memory cache from the external memory storage
One remaining area is captured to the memory cache from the external memory storage.Then, the method performs a motion compensation step, root
According to the reference zone for being stored in the memory cache compensation is moved for the image block.
Can be further understood by following detailed description and accompanying drawings on the advantages and spirit of the present invention.
Brief description of the drawings
Fig. 1 (A) is the work schedule example for representing present code device when compensation is moved；Fig. 1 (B) is to be used to
Represent work schedule example of the existing decoder when compensation is moved.
Fig. 2 is the encoder block figure in one embodiment of the invention.
Fig. 3 is the work schedule for representing encoder of the invention.
Fig. 4 is the corresponding relation example of a reference picture and reference zone.
Fig. 5 (A) and Fig. 5 (B) are the decoder block figure in one embodiment of the invention.
Fig. 6 is the image processor block diagram in one embodiment of the invention.
Fig. 7 and Fig. 8 are the image treatment method flow chart in embodiments of the invention.
Main element symbol description
T1~T9、T1’~T6’：Time section 200：Encoder
300：Decoder 400：Image processor
21、31、41：External memory storage 22,32,42：Memory cache
23、33、43：Motion compensation module 24,34,44：Judge module
25、35、45：Acquisition module 36：Amount of redundancy generation module
37：Analysis module 37R1 ~ R6：Block in reference picture
X1、X2：Reference zone S71 ~ S74, S81 ~ S86：Process step
Specific embodiment
An embodiment of the invention is the encoder 200 shown in Fig. 2, wherein being deposited comprising external memory storage 21, cache
Reservoir (cache) 22, motion compensation module 23, judge module 24 and acquisition module 25.In practical application, encoder 200 can
It is incorporated into various image processing systems or image playback equipment, can be also individually present.
External memory storage 21 is used to store the image block institute phase that will be processed or processing with motion compensation module 23
The reference picture of pass.More particularly, the reference picture is that each block of picture to be encoded produces motion-vector and redundancy according to this
The comparison basis of amount.Memory cache 22 is configured to temporarily store the number that small amount and motion compensation module 23 may need in a short time
According to.In practice, external memory storage 21 can using dynamic random access memory (dynamic random access memory,
DRAM) realize, memory cache 22 can using static RAM (static random access memory,
SRAM) realize.Compared to external memory storage 21, the hardware price of memory cache 22 may be higher, but accesses the speed of data
Degree is very fast.
As shown in Fig. 2 because the bit element flow of picture to be encoded is separately provided to motion compensation module 23 and judge module
24 so that motion compensation module 23 and judge module 24 can carry out image processing for different blocks simultaneously.In detail, it is assumed that move
Dynamic compensating module 23 sequentially moves compensation and compiles for two image blocks B1, B2 (being not depicted in schema) in picture to be encoded
Coded program, produces the amount of redundancy and motion-vector corresponding to image block B1, B2, as representing image block B1, B2 respectively
Coded data.When motion compensation module 23 moves compensation coded program to image block B1, judge module 24 can be sentenced
Motion-vectors of the disconnected image block B2 relative to reference picture.After judge module 24 produces the motion-vector of image block B2, pick
Modulus block 25 just according to this motion-vector, will in reference picture corresponding to the motion-vector a reference zone from external memory storage
21 capture to memory cache 22, are used when compensation coded program is moved to image block B2 for motion compensation module 23.
Fig. 3 is used to the work schedule of presentation code device 200.Time section T1 represents motion compensation module 23 to image block
B1 moves the time of compensation coded program.Time section T2 represents judge module 24 for producing the movement of image block B2
The time of vector.Assuming that the motion-vector that judge module 24 is produced points out that the reference zone for corresponding to image block B2 is distributed in ginseng
Examine the block R1 and block R2 in picture.Time section T3 is represented after acquisition module 25 sends out the requirement for capturing block R1 and is waited outer
The time that portion's memory 21 is responded.Time section T4 is represented and block R1 is transferred into memory cache 22 from external memory storage 21
Time.Time section T5 represent acquisition module 25 send out wait after the requirement for capturing block R2 external memory storage 21 to respond when
Between.Time section T6 represents the time that block R2 is transferred to memory cache 22 from external memory storage 21.Time section T7 and
T8 represents the time that motion compensation module 23 reads block R1 and R2 from memory cache 22 respectively.Time section T9 then represents shifting
Dynamic compensating module 23 moves the time of compensation coded program to image block B2.
As shown in figure 3, time section T2, T3 and time section T1 in this embodiment overlap completely, and time section T4
Also overlapped with time section T1 parts.Although each time section length in Fig. 3 is all only example, compared to Fig. 1 (A) Suo Shi
Conventional situation, due to encoder 200 be advanced by beginning from the time of the acquisition data of external memory storage 21, actually remain to compared with
Image block coded program is completed in the short time.It should be noted that, the initial time of time section T2 can be advanced to
The initial time of time section T1 is identical, even earlier than the initial time of time section T1.
In practice, judge module 24 can determine image block B2 by image block B2 and the reference picture is directly compared
Motion-vector, one can also be produced according to an at least adjacent block of image block B2 and/or at least one adjacent motion-vector
Motion-vector of the prediction motion-vector as image block B2.For example, if image block B1, B2 is adjacent to each other, judge
Module 24 can speculate the motion-vector of image block B2 according to the motion-vector of the image block B1 for having produced.Compared to direct
Compare image block and reference picture, by prediction come determining moving vectors it is typically more time saving, can further lift encoder
200 efficiency.In general, adjacent image block often has close motion-vector.Therefore, done according to adjacent motion-vector
For fundamentals of forecasting as a rule can obtain will not be too big with optimum value gap result.It should be noted that, judge module 24
The object of reference is not limited to the motion-vector of image block B1, and the motion-vector of prediction image block B2 is also by adjacent multiple shadows
As average after the weight of the motion-vector of block.Moreover, it is judged that module 24 also can be according to the known properties of image block B2 (for example
Whether it is object boundary in image) elasticity selection predicts the data of its motion-vector according to this.
From the above, the prediction motion-vector of image block B2 is generally not too large with the gap of an optimal motion-vector.
Therefore, motion compensation module 23 is designed to not consider the gap of predicted motion-vector and optimal motion-vector, also
Be no matter the motion-vector that judge module 24 is predicted why, motion compensation module 23 all according to this prediction motion-vector moved
Dynamic compensation coded program.In other words, motion compensation module 23 can be using image block B2 and the motion-vector for corresponding to the prediction
Reference zone between image difference as the amount of redundancy of image block B2, and produced according to the image difference and the motion-vector
Represent the coded data of image block B2.
In another embodiment, before motion compensation module 23 moves compensation coded program to image block B2,
Judge module 24 be designed to determine whether image block B2 relative to reference picture actual motion-vector (namely
Foregoing optimal motion-vector), and the motion-vector of comparison prediction and actual motion-vector.If prediction motion-vector is mobile with actual
The difference of vector is more than a preset threshold, and acquisition module 25 corresponds to real in further capturing reference picture from external memory storage 21
One actual reference zone of border motion-vector, uses for motion compensation module 23 when image block B2 is processed.Relatively, if in advance
Survey motion-vector and be less than or equal to the preset threshold with the difference of actual motion-vector, acquisition module 25 just need not again by outside
The acquisition data of memory 21, and motion compensation module 23 can be deposited directly according to corresponding to prediction motion-vector and having captured to cache
The reference zone of reservoir 22 is that image block B2 produces coded data.
It should be noted that, acquisition module 25 can control the actual reference zone and is subtracted to shifting directly from external memory storage 21
Dynamic compensating module 23, is not necessarily intended to by memory cache 22.In practical application, as long as judge module 24 has produced prediction to move
Moving vector, acquisition module 25 just can start to capture the reference zone for corresponding to the prediction motion-vector to memory cache 22,
Produced without waiting for actual motion-vector.Further, since needing to capture the machine of actual reference zone from external memory storage 21 again
Rate less, generally speaking, even if must again capture above-mentioned actual reference zone once in a while, the code efficiency of encoder 200 or excellent
In prior art.
As described above, adjacent image block often has close motion-vector.Therefore, adjacent image block is likely to
Adjacent reference zone can be corresponded to.Fig. 4 is the corresponding relation example of a reference picture and reference zone.In this example, shadow
Block R1, R2, R4, R5 in reference picture, the reference zone X2 of image block B2 are distributed in as the reference zone X1 of block B1
Then it is distributed across block R2, R3, R5, R6 in reference picture.Just outside is realized using a dynamic random access memory to deposit
For the situation of reservoir 21, the access features of dynamic random access memory are limited to, in order to allow the motion compensation module 23 can be with
Treatment image block B1, acquisition module 25 is likely to have to all completely capture block R1, R2, R4, R5 to cache deposit
Reservoir 22, and can not only capture reference zone X1.
In an embodiment, before the motion-vector produced according to judge module 24 captures a reference zone, mould is captured
Whether block 25 first determines whether there has been all or part of reference zone in memory cache 22.If in memory cache 22
There are whole reference zones, acquisition module 25 just no longer captures any data from external memory storage 21.If memory cache
There has been the reference zone of part in 22, acquisition module 25 is then reference zone to be not yet stored in soon only from external memory storage 21
One remaining area of access to memory 22 is captured to memory cache 22, is not repeated to capture the number being stored in memory cache 22
According to.In case of depicted in Fig. 4, in order to process image block B1, had in memory cache 22 block R1, R2, R4,
R5.If the motion-vector that judge module 24 is produced points out that the reference zone X2 of image block B2 is distributed in the block in reference picture
R2, R3, R5, R6, acquisition module 25 will only be that image block B2 captures block R3, R6 from external memory storage 21, without repeating to pick
Take block R2, R5.This way for not repeating acquisition data can further save the working time of encoder 200 and data are passed
Throughput rate.
It is according to another embodiment of the present invention the decoder 300 shown in Fig. 5 (A), wherein comprising external memory storage 31, fast
Access to memory 32, motion compensation module 33, judge module 34, acquisition module 35 and amount of redundancy generation module 36.In practical application
In, decoder 300 can be incorporated into various image processing systems or image playback equipment, can be also individually present.External storage
Device 31 is used to store the reference picture related to the image block that motion compensation module 33 will be processed or processing.It is clearer and more definite
Say that the reference picture is that each block of picture to be decoded produces the comparison base of motion-vector and amount of redundancy in coding side according to this in ground
Plinth.
As shown in Fig. 5 (A), the bit element flow of picture to be decoded is separately provided to judge module 34 and amount of redundancy generation module
36.Amount of redundancy generation module 36 is used to analyze the bit element flow, real with obtain image block use corresponding with coding side one
Image difference between the reference zone of border, that is, the image block amount of redundancy.Assuming that motion compensation module 33 is sequentially directed to treating
Two image blocks B1, B2 in decoded picture move compensation decoding program, and image block B1, B2 is reconstructed respectively.Moving
When dynamic compensating module 33 moves compensation decoding program to image block B1, judge module 34 is responsible for judging image block B2 phases
For the motion-vector of reference picture.After the motion-vector that judge module 34 produces image block B2, the just root of acquisition module 35
Motion-vector, the reference zone in reference picture corresponding to the motion-vector is captured to cache from external memory storage 31 accordingly
Memory 32, uses when compensation decoding program is moved to image block B2 for motion compensation module 33.Motion compensation module
The 33 image difference reconstructed image block B2 that can be drawn according to the reference zone and amount of redundancy generation module 36.
Similarly, judge module 34 can be according to an at least adjacent block of image block B2 and/or at least one adjacent movement
Vector produces one to predict motion-vector of the motion-vector as image block B2.Motion compensation module 33 is designed to not consider
The gap of prediction motion-vector and actual motion-vector, that is, no matter the prediction motion-vector produced by judge module 34 is
What, motion compensation module 33 all moves compensation decoding program according to this prediction motion-vector.Compared to shown in Fig. 1 (B)
Conventional situation, because decoder 300 is advanced by beginning from the time of the acquisition data of external memory storage 31, therefore can be when shorter
Interior completion image block decoding program.
Fig. 5 (B) illustrates another embodiment of decoder 300.In this embodiment, the amount of redundancy generation module in Fig. 5 (A)
36 are replaced by an analysis module 37.Analysis module 37 is used to analyze the bit element flow, to obtain an image block relative to coding
Hold the image difference and actual motion-vector between the actual reference zone for using.It is worth noting that, being obtained in analysis module 37
Before the actual motion-vector, acquisition module 35 has started to capture the reference zone to memory cache 32.In analysis mould
After block 37 obtains the actual motion-vector, judge module 34 compares the prediction motion-vector and the actual motion-vector.If prediction
Motion-vector is more than a preset threshold with the difference of actual motion-vector, and acquisition module 35 can be picked further from external memory storage 31
The actual reference zone is taken, is used when the image block is processed for motion compensation module 33.Relatively, if prediction motion-vector
Preset threshold is less than or equal to the difference of actual motion-vector, motion compensation module 33 is just directly used and captured to cache storage
The reference zone corresponding to prediction motion-vector of device 32 moves compensation decoding program.
In practical application, the previously described concept for not repeating to capture also can be applicable to shown in Fig. 5 (A) or Fig. 5 (B)
In decoder 300.
It is according to another embodiment of the present invention the image processor 400 shown in Fig. 6, wherein comprising external memory storage
41st, memory cache 42, motion compensation module 43, judge module 44 and acquisition module 45.In practical application, image processing dress
Putting 400 can be incorporated into various image processing systems or image playback equipment, can also be individually present.External memory storage 41 is used to
A storage reference picture related to an image block.Motion compensation module 43 is used to carry out a movement benefit for the image block
Repay program (such as a motion compensation coded program or a motion compensation decoding program).Judge module 44 is used to judge the image area
A motion-vector of the block relative to the reference picture.Acquisition module 45 is used to confirm correspond to the motion-vector in the reference picture
A reference zone, and judge whether there has been all or part of reference zone in memory cache 42.
If there not being any part of the reference zone in memory cache 42, acquisition module 45 can be whole by the reference zone
It is individual to be captured to memory cache 42 from external memory storage 41.If there being the reference zone of part in memory cache 42, pick
Modulus block 45 is only captured to fast the remaining area that the reference zone is not yet stored in memory cache 42 from external memory storage 41
Access to memory 42, uses for motion compensation module 43 when the image block is processed.Relatively, if having been deposited in memory cache 42
There are all parts of the reference zone, acquisition module 45 will not be captured again.By avoid repeat acquisition data, image processing
Device 400 can effectively save the working time.
It is according to another embodiment of the present invention a kind of image with unification external memory storage and memory cache running
Processing method, its flow chart is as shown in Figure 7.The external memory storage is used to store a reference picture related to an image block.
First, step S71 is to carry out a motion compensation program to image block before.Step S72 is to start to sentence before step S71 terminates
A motion-vector of the image block of breaking relative to the reference picture.After the completion of step S72, step S73 is performed, and this is joined
Examine and correspond to a reference zone of the motion-vector in picture and captured to the memory cache from the external memory storage.Step S74
It is then that the motion compensation program is carried out to the image block.Previously described when phase encoder 200, decoder 300 is introduced
Several circuit operation flow changes, can also be applied into the image treatment method depicted in Fig. 7, and its details is repeated no more.
It is according to another embodiment of the present invention a kind of image with unification external memory storage and memory cache running
Processing method, its flow chart is as shown in Figure 8.The external memory storage is used to store a reference picture related to an image block.
First, step S81 is to judge a motion-vector of the image block relative to the reference picture.Step S82 is then to confirm the ginseng
Examine the reference zone for corresponding to the motion-vector in picture.Whether step S83 is to judge had entirely in the memory cache
Portion or the partial reference zone.If the judged result of step S83 points out there is not the reference zone in the memory cache
Any part, step S84 will be performed, and store whole the acquisition to the cache of the reference zone with from the external memory storage
Device.If the judged result of step S83 points out the reference zone for having had part in the memory cache, step S85 will be by
Perform, capture the remaining area that the reference zone is not yet stored in the memory cache from the external memory storage with only fast to this
Access to memory.If the judged result of step S83 points out all parts for having had the reference zone in the memory cache, step
S86 is directly performed, and a movement benefit is carried out for the image block with according to the reference zone for being stored in the memory cache
Repay program.Motion compensation program in step S86 can be a motion compensation coded program or a motion compensation decoding program.
As described above, the above embodiment of the present invention proposes a kind of image processor and image treatment method, by right
When one image block imposes motion compensation program simultaneously for next image block it is advance from external memory storage by palpus reference number
According to capturing to memory cache, timesaving effect is reached.Additionally, image processor according to the above embodiment of the present invention and shadow
As processing method thereby can further lift treatment effect using the concept of the data for not repeating to have had in acquisition memory cache
Rate.
The above detailed description of preferred embodiments wishes to more to clearly describe feature of the invention and spirit, and not with
Above-mentioned disclosed preferred embodiment is any limitation as to scope of the invention.On the contrary, the purpose is to wish to cover
Various changes and tool equality are arranged in the category of the scope of the claims to be applied of the invention.
Claims (18)
1. a kind of motion compensation image processor, comprising：
One external memory storage, is used to store a reference picture related to an image block；
One memory cache；
One motion compensation module, compensation is moved to image block before being sequentially directed to one and the image block；
One judge module, judges that the image block is relative when the motion compensation module moves compensation to the preceding image block
In a motion-vector of the reference picture；And
One acquisition module, to when the motion compensation module moves compensation to the preceding image block, by the reference picture
In correspond to the motion-vector a reference zone captured to the memory cache from the external memory storage；
Wherein, the bit element flow of picture to be encoded is separately provided to the motion compensation module and the judge module so that the movement
When compensating module moves compensation to the preceding image block, the judge module can simultaneously the image block is carried out the movement to
Amount judges.
2. image processor as claimed in claim 1, it is characterised in that the judge module is according to the adjacent image block
The type information of one adjacent block or the adjacent motion vector prediction motion-vector.
3. image processor as claimed in claim 2, it is characterised in that the motion compensation module judge the image block with
An image difference between the reference zone, and the volume for representing the image block is produced according to the image difference and the motion-vector
Data after code.
4. image processor as claimed in claim 2, it is characterised in that also include：
One amount of redundancy generation module, be used to analyze the coded data corresponding to the image block with obtain the image block with
An image difference between one actual reference zone；
Wherein the motion compensation module rebuilds the image block according to the reference zone and the image difference.
5. image processor as claimed in claim 2, it is characterised in that enter to the image block in the motion compensation module
Before row motion compensation, the judge module also judges an actual motion-vector of the image block, and compares the motion-vector and be somebody's turn to do
Actual motion-vector；Before the actual motion-vector is produced, the acquisition module is to start to capture the reference zone to the cache
Memory；If the motion-vector is more than a preset threshold with the difference of the actual motion-vector, the acquisition module is deposited from outside this
Reservoir captures the actual reference zone corresponding to the actual motion-vector.
6. image processor as claimed in claim 2, it is characterised in that further include：
One analysis module, is used to analyze the coded data corresponding to the image block obtain the image block relative to one
One image difference of actual reference zone and an actual motion-vector；
Wherein before the analysis module obtains the actual motion-vector, the acquisition module is to start to capture the reference zone to this
Memory cache；The judge module also compares the prediction motion-vector and the actual motion-vector, when the prediction motion-vector with
The difference of the actual motion-vector is more than a preset threshold, and the acquisition module captures the actual reference area from the external memory storage
Domain.
7. image processor as claimed in claim 1, it is characterised in that the acquisition module joins this from the external memory storage
The remaining area that examination district domain is not stored in the memory cache is captured to the memory cache.
8. a kind of with the motion compensation image treatment method for unifying external memory storage and memory cache running, the external storage
Device stores a reference picture related to an image block, and the image treatment method is included：
A () moves compensation to image block before；
B () starts to judge a motion-vector of the image block relative to the reference picture before step (a) terminates；
(c) before step (a) terminates, will in the reference picture corresponding to the motion-vector a reference zone from the external storage
Device is captured to the memory cache；And
D () moves compensation to the image block.
9. image treatment method as claimed in claim 8, it is characterised in that step (b) is included according to the adjacent image block
An adjacent block type information or the adjacent motion vector prediction motion-vector.
10. image treatment method as claimed in claim 9, it is characterised in that step (d) is included：
Judge the image difference between the image block and the reference zone；And
The coded data for representing the image block is produced according to the image difference and the motion-vector.
11. image treatment methods as claimed in claim 9, it is characterised in that also included before step (d)：
Analysis corresponds to a coded data of the image block to obtain between the image block and an actual reference zone
Image difference；
Wherein step (d) is included and rebuilds the image block according to the reference zone and the image difference.
12. image treatment methods as claimed in claim 9, it is characterised in that advance of the image treatment method in step (d)
One step is included：
Judge the image block relative to the one of the reference picture actual motion-vector；
Compare the motion-vector and the actual motion-vector；And
If the motion-vector is more than a preset threshold with the difference of the actual motion-vector, is captured from the external memory storage and corresponded to
One actual reference zone of the actual motion-vector；
Wherein before the actual motion-vector is produced, step (c) has started to perform.
13. image treatment methods as claimed in claim 9, it is characterised in that included in taking a step forward for step (d)：
Analysis corresponds to a coded data of the image block to obtain the image block relative to an actual reference zone
One image difference and an actual motion-vector；
Compare the prediction motion-vector with the actual motion-vector to produce a comparative result；And
If the comparative result is more than a preset threshold, the reality corresponding to the actual motion-vector is captured from the external memory storage
Reference zone；
Wherein before the actual motion-vector is obtained, step (c) has started to perform.
14. image treatment methods as claimed in claim 8, it is characterised in that step (c) is included
The remaining area that the reference zone is not stored in the memory cache is captured to cache storage from the external memory storage
Device.
A kind of 15. motion compensation image processors, comprising：
One external memory storage, is used to store a reference picture related to an image block；
One memory cache；
One motion compensation module, is used to move compensation for the image block；
One judge module, the object boundary property according to an adjacent image block judges that the image block refers to picture relative to this
One motion-vector in face；And
One acquisition module, the reference zone for corresponding to the motion-vector is not stored in into the cache from the external memory storage
One remaining area of memory is captured to the memory cache.
16. image processors as claimed in claim 15, it is characterised in that the motion compensation module moves compensation and compiles
Code or motion compensation decoding.
A kind of 17. motion compensation image treatment methods with unification external memory storage and memory cache running, deposit outside this
Reservoir is used to store a reference picture related to an image block, and the image treatment method is included：
Object boundary property according to an adjacent image block judges a movement of the image block relative to the reference picture
Vector；
A residue of the memory cache will not be stored in corresponding to a reference zone of the motion-vector from the external memory storage
Region is captured to the memory cache；And
The reference zone according to the memory cache is stored in moves compensation for the image block.
18. image treatment methods as claimed in claim 17, it is characterised in that motion compensation is that motion compensation is encoded or mobile
Compensation decoding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210251213.2A CN103581504B (en) | 2012-07-20 | 2012-07-20 | Motion compensation image processor and image treatment method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210251213.2A CN103581504B (en) | 2012-07-20 | 2012-07-20 | Motion compensation image processor and image treatment method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103581504A CN103581504A (en) | 2014-02-12 |
| CN103581504B true CN103581504B (en) | 2017-06-20 |
Family
ID=50052353
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210251213.2A Expired - Fee Related CN103581504B (en) | 2012-07-20 | 2012-07-20 | Motion compensation image processor and image treatment method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103581504B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7965773B1 (en) * | 2005-06-30 | 2011-06-21 | Advanced Micro Devices, Inc. | Macroblock cache |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10134A (en) * | 1853-10-18 | Cutting boots | ||
| CN101345026B (en) * | 2007-07-10 | 2010-12-01 | 联詠科技股份有限公司 | Frame data buffering apparatus and related frame data acquisition method |
-
2012
- 2012-07-20 CN CN201210251213.2A patent/CN103581504B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7965773B1 (en) * | 2005-06-30 | 2011-06-21 | Advanced Micro Devices, Inc. | Macroblock cache |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103581504A (en) | 2014-02-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106251392A (en) | For the method and apparatus performing to interweave | |
| CN101115205A (en) | Method and apparatus for processing video stream | |
| CN102763136A (en) | Method and apparatus for providing multi-threaded video decoding | |
| CN105681807A (en) | Method and device for calculating sub pixel motion vector based on H264 protocol | |
| TWI517677B (en) | Systems, methods, and computer program products for a video encoding pipeline | |
| CN107820091B (en) | Picture processing method and system and picture processing equipment | |
| CN104603834A (en) | Methods and systems for multimedia data processing | |
| WO2009110670A1 (en) | Encoding system using motion estimation and encoding method using motion estimation | |
| TWI441091B (en) | Method for performing image signal processing with aid of a graphics processing unit and apparatus for performing image signal processing | |
| CN102113327B (en) | Image encoding device, method, and integrated circuit | |
| Jiang et al. | Block-Based Parallel Intra Prediction Scheme for HEVC. | |
| Enfedaque et al. | GPU implementation of bitplane coding with parallel coefficient processing for high performance image compression | |
| TWI601075B (en) | Motion compensation image processing apparatus and image processing method | |
| CN1112654C (en) | Image processor | |
| US20110255596A1 (en) | Frame rate up conversion system and method | |
| CN103581504B (en) | Motion compensation image processor and image treatment method | |
| CN112203091B (en) | Motion vector prediction method, system and computer medium based on quadratic polynomial | |
| CN109151476A (en) | A kind of reference frame generating method and device based on bi-directional predicted B frame image | |
| CN105100799A (en) | Method for reducing intraframe coding time delay in HEVC encoder | |
| Seitner et al. | Evaluation of data-parallel H. 264 decoding approaches for strongly resource-restricted architectures | |
| CN101194513B (en) | Up-down alternative motion vector | |
| CN100496126C (en) | Image coder and coding method | |
| Doan et al. | Multi-asip based parallel and scalable implementation of motion estimation kernel for high definition videos | |
| JPH11346369A (en) | Motion vector detection circuit | |
| CN1110209C (en) | Device for estimation of movement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170620 Termination date: 20190720 |