CN105224290B - A kind of picture load equalization processing method and device - Google Patents
A kind of picture load equalization processing method and device Download PDFInfo
- Publication number
- CN105224290B CN105224290B CN201510618811.2A CN201510618811A CN105224290B CN 105224290 B CN105224290 B CN 105224290B CN 201510618811 A CN201510618811 A CN 201510618811A CN 105224290 B CN105224290 B CN 105224290B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- image block
- textural characteristics
- col
- 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)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of picture load equalization processing method and device, method comprises the following steps:Original image f (x, y) is divided into multiple images block along vertical and horizontal;The textural characteristics of two image blocks in left and right of longitudinal boundary are calculated, mobile longitudinal boundary makes the textural characteristics similarity between two image blocks of longitudinal boundary or so maximum;The textural characteristics of two image blocks up and down of horizontal boundary are calculated, mobile horizontal boundary makes the textural characteristics similarity above and below longitudinal boundary between two image blocks maximum;The corresponding final image block of the final horizontal boundary and longitudinal boundary of determination is respectively allocated to different computing units and carries out parallel processing;The textural characteristics X (i) of image block is calculated by following algorithm:μcol(i)=∑ xm,n(i)/(Wp×Hp), X (i)=∑ | xm,n(i)‑μcol(i)|.This method can realize the load balancing between different parallel processing elements, reduce power consumption.
Description
【Technical field】
The present invention relates to a kind of picture load equalization processing method and device.
【Background technology】
The development of computing technique vision technique, the proposition of video coding technique of new generation, let us enjoys high-resolution
The vision grand banquet come with ultrahigh resolution image and video tape, but this also causes the calculating in image procossing and Video coding
Complexity increases substantially, and is that parallel super-resolution image/video processing brings huge challenge in real time, while parallel processing architecture
The problem of load imbalance, power consumption penalty increase may also be brought, it is how that the parallel processing strategy of real-time high-efficiency and load is equal
Weighing apparatus technological perfectionism combines the development trend for having become computer vision and digital image processing field.
Either image procossing or Video coding, the important composition part of its concurrent technique is the method for piecemeal.Mesh
Preceding method of partition easily causes the load imbalance between different parallel processing elements.
【The content of the invention】
In order to solve the deficiencies in the prior art, the invention provides a kind of picture load equalization processing method and device, from
And make it that the load between different parallel processing elements is as balanced as possible.
A kind of picture load equalization processing method, comprises the following steps:
S1, multiple images block is divided into by original image f (x, y) along vertical and horizontal;
S2, calculates the textural characteristics of two image blocks in left and right of longitudinal boundary, and the mobile longitudinal boundary makes the longitudinal direction
Textural characteristics similarity between the image block of two, border or so is maximum;
S3, calculates the textural characteristics of two image blocks up and down of horizontal boundary, and the mobile horizontal boundary makes the transverse direction
Textural characteristics similarity above and below border between two image blocks is maximum;
S4, different meters are respectively allocated to by the corresponding final image block of the final horizontal boundary and longitudinal boundary of determination
Calculate unit and carry out parallel processing;
Wherein, the textural characteristics X (i) of image block is calculated by following algorithm in step S2 and S3:
μcol(i)=∑ xm,n(i)/(Wp×Hp);
X (i)=∑ | xm,n(i)-μcol(i)|;
Wherein xm,n(i) pixel value of pixel in i-th piece of image block, W are representedpRepresent the horizontal width of initial image block
Degree, HPRepresent the longitudinal direction height of image block.
In one embodiment, in step S2 and S3, determine that the texture between two image blocks is special as follows
Levy similarity maximum:
S31, calculates the textural characteristics X (i) of i-th piece of adjacent image block and the textural characteristics X (i+ of i+1 block image block
1);
S32, compares between the textural characteristics X (i) of i-th piece of image block and the textural characteristics X (i+1) of i+1 block image block
Size, if X (i) is larger using i+1 block image block to the direction of i-th piece of image block as correspondence border moving direction,
I-th piece of image block to the direction of i+1 block image block is regard as the moving direction for corresponding to border if X (i+1) is larger;
S33, along moving direction using elementary cell as step-length moving boundary, two image blocks of cycle calculations boundaries on either side
Textural characteristics discrepancy delta, until textural characteristics discrepancy delta reaches minimum value;Wherein, Δ=| X (i+1)-X (i) |;
S34, by the border between the minimum corresponding i-th piece of image block of textural characteristics discrepancy delta and i+1 block image block
It is used as final border.
In one embodiment, if the moving direction on border to the left, is moved after elementary cell Δ W:
X (i)=∑ | xm,n(i)-μcol(i)|-∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|+∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
In one embodiment, if the moving direction on border to the right, is moved after elementary cell Δ W:
X (i)=∑ | xm,n(i)-μcol(i)|+∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|-∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
Present invention also offers a kind of picture load equilibrium treatment device, including such as lower unit:
First module, for original image f (x, y) to be divided into multiple images block along vertical and horizontal;
Second unit, the textural characteristics of two image blocks in left and right for calculating longitudinal boundary, the mobile longitudinal boundary
Make the textural characteristics similarity between two image blocks of described longitudinal boundary or so maximum;
Third unit, the textural characteristics of two image blocks up and down for calculating horizontal boundary, the mobile horizontal boundary
Make the textural characteristics similarity above and below the horizontal boundary between two image blocks maximum;
Unit the 4th, for the corresponding final image block of the final horizontal boundary and longitudinal boundary of determination to be distributed respectively
Parallel processing is carried out to different computing units;
Wherein, the textural characteristics X (i) of image block is calculated by following algorithm in second unit and third unit:
μcol(i)=∑ xm,n(i)/(Wp×Hp);
X (i)=∑ | xm,n(i)-μcol(i)|;
Wherein xm,n(i) pixel value of pixel in i-th piece of image block, W are representedpRepresent the horizontal width of initial image block
Degree, HPRepresent the longitudinal direction height of image block.
In one embodiment, in second unit and third unit, as follows between two image blocks of determination
Textural characteristics similarity it is maximum:
Calculate the textural characteristics X (i) of i-th piece of adjacent image block and the textural characteristics X (i+1) of i+1 block image block;
Compare big between the textural characteristics X (i) of i-th piece of image block and the textural characteristics X (i+1) of i+1 block image block
It is small, using i+1 block image block to the direction of i-th piece of image block as the moving direction for corresponding to border if X (i) is larger, if X (i
+ 1) it is larger then to regard i-th piece of image block to the direction of i+1 block image block as the moving direction for corresponding to border;
Along moving direction using elementary cell as step-length moving boundary, the texture of two image blocks of cycle calculations boundaries on either side
Feature difference Δ, until textural characteristics discrepancy delta reaches minimum value;Wherein, Δ=| X (i+1)-X (i) |;
Using the border between the minimum corresponding i-th piece of image block of textural characteristics discrepancy delta and i+1 block image block as
Final border.
In one embodiment, if the moving direction on border to the left, is moved after elementary cell Δ W:
X (i)=∑ | xm,n(i)-μcol(i)|-∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|+∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
In one embodiment, if the moving direction on border to the right, is moved after elementary cell Δ W:
X (i)=∑ | xm,n(i)-μcol(i)|+∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|-∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
The present invention determines Boundary Moving direction according to initial adjacent image block textural characteristics difference size, and then according to making
The principle moving boundary that adjacent two image blocks textural characteristics difference reduces is obtained, once there is the textural characteristics phase between adjacent image block
Then stop immediately like property reduction, the algorithm is effectively guaranteed adjacent parallel block and possesses similar grain feature, while reducing meter
Calculate complexity.
This method can realize the load balancing between different parallel processing elements, reduce power consumption, while meeting parallel
Degree demand, realizes efficient, real-time image Parallel Processing, and parallel efficiency is optimal, will cause computer vision and image procossing
Each research field is benefited extensively.
【Brief description of the drawings】
Fig. 1 is the flow chart of the picture load equalization processing method of an embodiment of the present invention;
Fig. 2 is the Boundary Moving schematic diagram of the image block of an embodiment of the present invention.
【Embodiment】
The preferred embodiment to invention is described in further detail below.
As illustrated in fig. 1 and 2, a kind of picture load equalization processing method of embodiment, comprises the following steps:
S1:Input original image, original image can be as video camera, camera collection gained, or pass through calculating
Video image obtained by machine instrument, data format is not limited.
S2:By original image f0It is evenly dividing as M × N rectangle original blocks, the image number of blocks of vertical and horizontal is respectively M
And N, the number of pixels of every piece of elementary cell is n × n block of pixels, horizontal to institute divided block longitudinal boundary i numberings from 1 to M-1
To border j numberings from 1 to N-1, it is 1 that initial value is assigned respectively;
S3:Longitudinal boundary i positions are determined, the adjacent piecemeal textural characteristics difference in left and right is calculated | X (i)-X (i+1) |, block texture
Feature uses the difference absolute value and expression of each point pixel value and average in block based on K-means algorithms, and calculation criterion is as follows:
μcol(i)=∑ xm,n(i)/(Wp×Hp)
X (i)=∑ | xm,n(i)-μcol(i)|
Wherein xm,n(i) pixel value of each pixel in i-th piece, W are representedpThe horizontal width of initial image block is represented,
And HPLongitudinal picture altitude is represented, X (i) values are bigger, show that texture is more complicated.
Calculate current border it is motionless and respectively to the left/feature difference of a basic step-length (n) is moved right, if current location
Difference is minimum then directly to be divided, and is otherwise defined as Boundary Moving direction according to minimum value direction.(sign is moved to right initialization k to the left
Dynamic step-length, using CTU as digit) it is equal to 0, adjacent two modules textural characteristics difference when k is 0,1, -1 is calculated respectively | X
(i)-X(i+1)|k, specific computational methods compare three and find out minimum value referring to S4, determine Boundary Moving direction.
S4:Class K-means cluster features are taken, the assessment of image texture characteristic difference between parallel block is realized, figure is completed
The division of parallel block as in.
When determining i-th border in image, using basic coding unit n × n as unit, side is moved respectively to the left or to the right
Boundary, traversal search k steps, the texture that i-th and i+1 blocks are calculated respectively characterizes X (i) and X (i+1), calculate sign this adjacent two
The absolute value of the feature difference of module | X (i)-X (i+1) |k, when | X (i)-X (i+1) |kLess than minimum M in, the value is given to
Min, k continue cycle criterion after adding 1;Otherwise image division is carried out by current location.Whole process also needs to judge whether arrival figure
As the adjacent i-1 and i+1 bars in border or left and right divide border, stop dividing if reaching border, otherwise continue cycling through.
To the left with move right border when pixel mean μ (i) it is different with textural characteristics X (i) computational methods, wherein to the left
Calculation criterion it is as follows:
X (i)=∑ | xm,n(i)-μcol(i)|-∑|ym,n(i)-μcol(i)|
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|+∑|ym,n(i+1)-μcol(i+1)|
Move right border when, calculation criterion is as follows:
X (i)=∑ | xm,n(i)-μcol(i)|+∑|ym,n(i)-μcol(i)|
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|-∑|ym,n(i+1)-μcol(i+1)|
X thereinm,n(i) i-th piece of pixel during expression original uniform is divided, and ym,n(i) i-th piece is represented because of Boundary Moving
And change the pixel of position, the width of Δ W changes.
S5:Judge whether that the parallel block for completing longitudinal direction is divided by i, if i adds 1 and is recycled to S3 when i is less than M, otherwise carry out
The determination of horizontal boundary, its flow is with longitudinal direction, and process is similar with S3, S4.
Above content is to combine specific preferred embodiment further description made for the present invention, it is impossible to assert
The specific implementation of the present invention is confined to these explanations.For general technical staff of the technical field of the invention,
On the premise of not departing from present inventive concept, some simple deduction or replace can also be made, should all be considered as belonging to the present invention by
The scope of patent protection that the claims submitted are determined.
Claims (6)
1. a kind of picture load equalization processing method, it is characterized in that, comprise the following steps:
S1, multiple images block is divided into by original image f (x, y) along vertical and horizontal;
S2, calculates the textural characteristics of two image blocks in left and right of longitudinal boundary, and the mobile longitudinal boundary makes the longitudinal boundary
Textural characteristics similarity between the image block of left and right two is maximum;
S3, calculates the textural characteristics of two image blocks up and down of horizontal boundary, and the mobile horizontal boundary makes the horizontal boundary
Textural characteristics similarity up and down between two image blocks is maximum;
S4, different calculating lists are respectively allocated to by the corresponding final image block of the final horizontal boundary and longitudinal boundary of determination
Member carries out parallel processing;
Wherein, the textural characteristics X (i) of image block is calculated by following algorithm in step S2 and S3:
μcol(i)=∑ xm,n(i)/(Wp×Hp);
X (i)=∑ | xm,n(i)-μcol(i)|;
Wherein xm,n(i) pixel value of pixel in i-th piece of image block, W are representedpRepresent the horizontal width of initial image block, HP
Represent the longitudinal direction height of image block;
In step S2 and S3, determine that the textural characteristics similarity between two image blocks is maximum as follows:
S31, calculates the textural characteristics X (i) of i-th piece of adjacent image block and the textural characteristics X (i+1) of i+1 block image block;
S32, compares big between the textural characteristics X (i) of i-th piece of image block and the textural characteristics X (i+1) of i+1 block image block
It is small, using i+1 block image block to the direction of i-th piece of image block as the moving direction for corresponding to border if X (i) is larger, if X (i
+ 1) it is larger then to regard i-th piece of image block to the direction of i+1 block image block as the moving direction for corresponding to border;
S33, along moving direction using elementary cell as step-length moving boundary, the texture of two image blocks of cycle calculations boundaries on either side
Feature difference Δ, until textural characteristics discrepancy delta reaches minimum value;Wherein, Δ=| X (i+1)-X (i) |;
S34, using the border between the minimum corresponding i-th piece of image block of textural characteristics discrepancy delta and i+1 block image block as
Final border.
2. picture load equalization processing method as claimed in claim 1, it is characterized in that, if the moving direction on border is to the left, move
After dynamic elementary cell Δ W:
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>-</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>+</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
X (i)=∑ | xm,n(i)-μcol(i)|-∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|+∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
3. picture load equalization processing method as claimed in claim 1, it is characterized in that, if the moving direction on border is to the right, move
After dynamic elementary cell Δ W:
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>+</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>-</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
1
X (i)=∑ | xm,n(i)-μcol(i)|+∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|-∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
4. a kind of picture load equilibrium treatment device, it is characterized in that, including such as lower unit:
First module, for original image f (x, y) to be divided into multiple images block along vertical and horizontal;
Second unit, the textural characteristics of two image blocks in left and right for calculating longitudinal boundary, the mobile longitudinal boundary makes institute
The textural characteristics similarity stated between two image blocks of longitudinal boundary or so is maximum;
Third unit, the textural characteristics of two image blocks up and down for calculating horizontal boundary, the mobile horizontal boundary makes institute
The textural characteristics similarity stated above and below horizontal boundary between two image blocks is maximum;
Unit the 4th, for the corresponding final image block of the final horizontal boundary and longitudinal boundary of determination to be respectively allocated to not
Same computing unit carries out parallel processing;
Wherein, the textural characteristics X (i) of image block is calculated by following algorithm in second unit and third unit:
μcol(i)=∑ xm,n(i)/(Wp×Hp);
X (i)=∑ | xm,n(i)-μcol(i)|;
Wherein xm,n(i) pixel value of pixel in i-th piece of image block, W are representedpRepresent the horizontal width of initial image block, HP
Represent the longitudinal direction height of image block;
In second unit and third unit, as follows determine two image blocks between textural characteristics similarity most
Greatly:
Calculate the textural characteristics X (i) of i-th piece of adjacent image block and the textural characteristics X (i+1) of i+1 block image block;
Compare the size between the textural characteristics X (i) of i-th piece of image block and the textural characteristics X (i+1) of i+1 block image block, if
X (i) it is larger then using i+1 block image block to the direction of i-th piece of image block as correspondence border moving direction, if X (i+1) compared with
Moving direction big then that i-th piece of image block to the direction of i+1 block image block is used as to correspondingly border;
Along moving direction using elementary cell as step-length moving boundary, the textural characteristics of two image blocks of cycle calculations boundaries on either side
Discrepancy delta, until textural characteristics discrepancy delta reaches minimum value;Wherein, Δ=| X (i+1)-X (i) |;
Using the border between the minimum corresponding i-th piece of image block of textural characteristics discrepancy delta and i+1 block image block as final
Border.
5. picture load equilibrium treatment device as claimed in claim 4, it is characterized in that, if the moving direction on border is to the left, move
After dynamic elementary cell Δ W:
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>-</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>+</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
X (i)=∑ | xm,n(i)-μcol(i)|-∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|+∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
6. picture load equilibrium treatment device as claimed in claim 4, it is characterized in that, if the moving direction on border is to the right, move
After dynamic elementary cell Δ W:
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>+</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
<mrow>
<msub>
<mi>&mu;</mi>
<mrow>
<mi>c</mi>
<mi>o</mi>
<mi>l</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;x</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>&Sigma;y</mi>
<mrow>
<mi>m</mi>
<mo>,</mo>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>W</mi>
<mi>p</mi>
</msub>
<mo>-</mo>
<mi>&Delta;</mi>
<mi>W</mi>
<mo>)</mo>
<msub>
<mi>H</mi>
<mi>p</mi>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
X (i)=∑ | xm,n(i)-μcol(i)|+∑|ym,n(i)-μcol(i)|;
X (i+1)=∑ | xm,n(i+1)-μcol(i+1)|-∑|ym,n(i+1)-μcol(i+1)|;
Wherein, ym,n(i) pixel for the position that i-th piece of image block changes by Boundary Moving is represented.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510618811.2A CN105224290B (en) | 2015-09-24 | 2015-09-24 | A kind of picture load equalization processing method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510618811.2A CN105224290B (en) | 2015-09-24 | 2015-09-24 | A kind of picture load equalization processing method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105224290A CN105224290A (en) | 2016-01-06 |
CN105224290B true CN105224290B (en) | 2017-11-07 |
Family
ID=54993292
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510618811.2A Active CN105224290B (en) | 2015-09-24 | 2015-09-24 | A kind of picture load equalization processing method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105224290B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104142810A (en) * | 2014-07-14 | 2014-11-12 | 中国南方电网有限责任公司电网技术研究中心 | Parallel method for forming nodal admittance matrix |
US9035678B2 (en) * | 2011-02-16 | 2015-05-19 | Broadcom Corporation | Method and apparatus for reducing power consumption in a digital circuit by controlling the clock |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004102989A (en) * | 1998-08-31 | 2004-04-02 | Seiko Epson Corp | Image processing system, image processing method, and medium recording image processing module |
-
2015
- 2015-09-24 CN CN201510618811.2A patent/CN105224290B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9035678B2 (en) * | 2011-02-16 | 2015-05-19 | Broadcom Corporation | Method and apparatus for reducing power consumption in a digital circuit by controlling the clock |
CN104142810A (en) * | 2014-07-14 | 2014-11-12 | 中国南方电网有限责任公司电网技术研究中心 | Parallel method for forming nodal admittance matrix |
Also Published As
Publication number | Publication date |
---|---|
CN105224290A (en) | 2016-01-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110533164B (en) | Winograd convolution splitting method for convolution neural network accelerator | |
CN108510485B (en) | Non-reference image quality evaluation method based on convolutional neural network | |
EP4024280A1 (en) | Data processing method and apparatus, computer equipment, and storage medium | |
US20210374511A1 (en) | Data processing method, device, computer equipment and storage medium | |
CN107480770A (en) | The adjustable neutral net for quantifying bit wide quantifies the method and device with compression | |
CN112435282B (en) | Real-time binocular stereo matching method based on self-adaptive candidate parallax prediction network | |
CN111882002A (en) | MSF-AM-based low-illumination target detection method | |
CN107229598A (en) | A kind of low power consumption voltage towards convolutional neural networks is adjustable convolution computing module | |
CN110533022B (en) | Target detection method, system, device and storage medium | |
CN103402045A (en) | Image de-spin and stabilization method based on subarea matching and affine model | |
CN102243711A (en) | Neighbor embedding-based image super-resolution reconstruction method | |
CN112257844B (en) | Convolutional neural network accelerator based on mixed precision configuration and implementation method thereof | |
CN103281537B (en) | A kind of dynamic range of images compression method and device | |
CN109949221A (en) | A kind of image processing method and electronic equipment | |
CN111598918A (en) | Video image stabilization motion estimation method based on reference frame optimization and foreground and background separation | |
CN108596992B (en) | Rapid real-time lip gloss makeup method | |
CN111091183B (en) | Neural network acceleration system and method | |
CN111626160A (en) | Face detection method under angle change based on regional progressive calibration network | |
CN103024247A (en) | Electronic image stabilization method based on improved block matching | |
CN108171651B (en) | Image alignment method based on multi-model geometric fitting and layered homography transformation | |
CN105224290B (en) | A kind of picture load equalization processing method and device | |
CN117237190B (en) | Lightweight image super-resolution reconstruction system and method for edge mobile equipment | |
CN111488839B (en) | Target detection method and target detection system | |
CN113630375A (en) | Compression apparatus and method using parameters of quadtree method | |
Ling et al. | Hardware accelerator for an accurate local stereo matching algorithm using binary neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information |
Address after: 518055 Guangdong city of Shenzhen province Nanshan District Xili of Tsinghua Applicant after: Graduate School at Shenzhen, Tsinghua University Address before: 518000 Guangdong city in Shenzhen Province, Nanshan District City Xili Shenzhen Tsinghua Campus of Tsinghua University Applicant before: Graduate School at Shenzhen, Tsinghua University |
|
COR | Change of bibliographic data | ||
GR01 | Patent grant | ||
GR01 | Patent grant |