CN107194867A - A kind of stingy picture synthetic method based on CUDA - Google Patents
A kind of stingy picture synthetic method based on CUDA Download PDFInfo
- Publication number
- CN107194867A CN107194867A CN201710336870.XA CN201710336870A CN107194867A CN 107194867 A CN107194867 A CN 107194867A CN 201710336870 A CN201710336870 A CN 201710336870A CN 107194867 A CN107194867 A CN 107194867A
- Authority
- CN
- China
- Prior art keywords
- mrow
- mtd
- shade
- point
- image
- 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.)
- Pending
Links
- HPTJABJPZMULFH-UHFFFAOYSA-N 12-[(Cyclohexylcarbamoyl)amino]dodecanoic acid Chemical compound OC(=O)CCCCCCCCCCCNC(=O)NC1CCCCC1 HPTJABJPZMULFH-UHFFFAOYSA-N 0.000 title claims abstract description 12
- 238000010189 synthetic method Methods 0.000 title claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000006748 scratching Methods 0.000 claims abstract description 10
- 230000002393 scratching effect Effects 0.000 claims abstract description 10
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 8
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 8
- 238000000034 method Methods 0.000 claims abstract description 7
- 239000007787 solid Substances 0.000 claims abstract description 5
- 238000012360 testing method Methods 0.000 claims abstract description 3
- 230000015654 memory Effects 0.000 claims description 14
- 230000000694 effects Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 9
- 230000001360 synchronised effect Effects 0.000 claims description 4
- 230000000875 corresponding effect Effects 0.000 claims description 3
- 230000002596 correlated effect Effects 0.000 claims description 2
- 238000003860 storage Methods 0.000 claims description 2
- 230000004927 fusion Effects 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 8
- 238000011160 research Methods 0.000 description 8
- 238000011161 development Methods 0.000 description 4
- 230000018109 developmental process Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 230000035800 maturation Effects 0.000 description 2
- 238000010422 painting Methods 0.000 description 2
- BASFCYQUMIYNBI-UHFFFAOYSA-N platinum Chemical compound [Pt] BASFCYQUMIYNBI-UHFFFAOYSA-N 0.000 description 2
- WKBPZYKAUNRMKP-UHFFFAOYSA-N 1-[2-(2,4-dichlorophenyl)pentyl]1,2,4-triazole Chemical compound C=1C=C(Cl)C=C(Cl)C=1C(CCC)CN1C=NC=N1 WKBPZYKAUNRMKP-UHFFFAOYSA-N 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 229910052697 platinum Inorganic materials 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000011031 topaz Substances 0.000 description 1
- 229910052853 topaz Inorganic materials 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/14—Transformations for image registration, e.g. adjusting or mapping for alignment of images
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of stingy picture synthetic method based on CUDA.First, spilling Balance Treatment is carried out to foreground image non-green region using color balance operator；Then a shade is generated using basic shade operator, scratches and remove solid background；Next basic shade is strengthened using details shade operator.After above-mentioned processing, final shade is formed, is merged with background image, the image after output synthesis.System make use of GPU to accelerate in stingy picture and synthesis phase, further increase splicing speed.Test result indicates that, this method can effectively realize the real-time image scratching of video, and frame per second reaches 10 frames, and the requirement of real time of video keying can be more disclosure satisfy that compared to traditional CPU versions.
Description
Technical field
The invention discloses a kind of stingy picture synthetic method based on CUDA, belong to digital filming technical field.
Background technology
With the maturation that continues to develop of digital filming technology, movie and television play definition is improved constantly, live telecast also gradually to
High definition even super clear image quality is drawn close.In the case, also improved constantly for the requirement scratched as efficiency.How to ensure
The original image quality of video and good scratch are scratched as speed is the focus studied instantly as being improved on the premise of effect.Therefore algorithm is improved
Operational efficiency, lift the speed of service using existing hardware has important reality should for industries such as production of film and TV and live telecasts
With value.
The maturation developed with technology, present GPU is widely applied in universal parallel calculating processing, profit
The optimization design of parallel computation is carried out with GPU, GPU computation capabilities are given full play to, ever-increasing picture can be not only met
Plain processing requirement, more can save development cost by ripe PC hardware platform advantages.PC platform softwares are easy to modification simultaneously
The advantage of upgrading, enables to user using left-hand seat is easy to, while more productive values can also be produced.
Domestic production of film and TV industry fast development in recent years, therefore for the demand also more and more higher of correlation technique, it is domestic
Scholar also joins in the research of association area one after another, and achieves certain achievement.
The human relations of horse platinum are analyzed using Motion Control systems and later stage composite software After Effects as case,
Mainly study the application of its interactive photographic control system of high-tech in film shooting making.Analyze how it is solved in film
Real shooting condition does not allow or the dangerous high and high picture of capital cost this problem in manufacturing process.[] this
Research describes the current situation of domestic and international film special efficacy, describes Motion Control systems answering in film special efficacy
With having carried out brief introduction as correlation theory to scratching, analyzed and scratch as the domestic bottleneck problem of technology.This research is existing for understanding
The problem of stage domestic and international film special effect making state of development and domestic correlation technique are present tool has very great help, but this grinds
Study carefully and rest on theoretic, do not applied correlation theory, have some limitations.
The prosperous good general CUDA technologies in river are applied to physics field, and it is first to the parallel architecture of graphics processor and CUDA skills
The related programming model of art is described.Then graphics processor speed-up computation target and some electromagnetism of environmental correclation are utilized
The problem of scattering field.[] the research is more deep for the development of CUDA technologies and the parallel architecture research of graphics processor,
Introduction is very detailed, has certain reference and evocation to this research.But CUDA technologies are applied to calculate by researcher
Electromagnetic scattering problems in physics, rather than scratch as in processing, stingy picture processing and EM scatter model are calculated also in the presence of necessarily
Difference.This research is used for reference it and scratched to the theoretical research of CUDA technologies, being applied to as in processing.
Deng Shuan, Li Leimin, topaz is clear et al. to be directed to video keying algorithm traditional under complex scene to target object and the back of the body
The problem of image segmentation that scape texture is similar or obscure boundary is clear is difficult, it is proposed that a kind of view-based access control model sensor and laser radar
The video real-time image scratching algorithm that information is blended.The algorithm obtains area-of-interest depth from original laser radar cloud data
Information, and the stingy picture algorithm of improved spectrum is fused to as priori, create area-of-interest depth and scratch as Laplacian Matrix,
Drawn and scratched as result by the optimal iteration of clustering algorithm, and with Steerable filter device to scratching as result is post-processed.Through experiment
Operational efficiency can be improved while obtaining good effect by showing this algorithm.The video that this algorithm is directed under complex environment enters
Row is scratched has certain advantage as operating, but is due to that it there are certain requirements for hardware device tool, so that in the presence of certain office
It is sex-limited.
A kind of image processing method based on GPU of Chen Shifu, Zhang Shu, Zhang Le et al. design.Main operational steps are：First
Pending image is obtained, a figure layer of the pending image is copied the video memory of graphics processing unit to from Installed System Memory
In；Then an image procossing filter is started, when filter params are adjusted, in the current visible region of the figure layer
Thumbnail is calculated and obtains regulation result in real time；Finally entered using the thumbnail of each figure layer of the pending image
The painting canvas to be rendered into painting canvas, and is shown to window by row by the supreme mixing in bottom.This method is advantageous in that can be one
Determine to reduce occupancy of the system for video memory in degree, and rendering result can be shown in real time, but be limited in that its processing speed
Lifting is not fairly obvious.
The content of the invention
The technical solution adopted by the present invention is a kind of stingy picture synthetic method based on CUDA, first, is calculated using color balance
Son carries out spilling Balance Treatment to foreground image non-green region；Then a shade is generated using basic shade operator, scratches and go
Solid background；Next basic shade is strengthened using details shade operator.After above-mentioned processing, final hide is formed
Cover, is merged with background image, the image after output synthesis.
It make use of GPU to accelerate in stingy picture and synthesis phase, further increase splicing speed.Test result indicates that, the party
Method can effectively realize the real-time image scratching of video, and frame per second reaches 10 frames, and video can be more disclosure satisfy that compared to traditional CPU versions
Scratch the requirement of real time of picture.
Overflow balanced algorithm：
Balanced algorithm is overflowed to for G in foreground picture>R point, there is nG=R, so drops G values in the point of green
Low, function representation is：
It is improved to overflowing balanced algorithm：For G<(B+R)/2, i.e. region 3, pixels illustrated point is not green, then enters
Row reservation process；For G>(B+R)/2, i.e. region 1 and 2, pixels illustrated point G values are too high, then are balanced processing, take new G=
(R+B)/2, it is changed into (R+B)/2 equivalent to R in (1) formula.Its function expression is：
Next it is improved again, introduces parameter a, span is 0-1.
Using B=R straight lines as symmetry axis, parameter a determines ∠ ACD size, as a=1, ACD=0 ° of angle, as a=0,
∠ ACD=90 °, i.e. ∠ ACD=(1-a) * 90, it can thus be appreciated that straight line AC slopes, obtain AC linear equations, you can distinguishable region 1.Ginseng
Number a changes the size with reservation region equivalent to the value for changing F (G, B, R).
Region 1：F (G, B, R)=a*R+ (1-a) * B；
Region 2：F (G, B, R)=(1-a) * R+a*B；
Region 3：F (G, B, R)=G；
It is identical with (2) formula as a=0.5, therefore function expression is changed into：
For convenience of being calculated, BR plans above are switched into BG plans.Abscissa represents G values (0-255) in way, indulges
Coordinate representation B values (0-255), represent a tangent plane in RGB color.
Input (R, G, B) is source images current pixel point, and a is the parameter spill for overflowing balance, can according to the R values of input
To find point C (R, R) in GB plane right-angle coordinate；According to a obtain point A (R* (1-a), 0), point B ((255-R) * a+R,
255) a is ratio, makes A points can only be in 0-R scopes, B points can only be in the range of R-255；Two can be made by three above point directly
Line Line1：B=(G-R* (1-a))/a, Line2：B=(G-R)/a+R；According to Line1, Line2, Line3 (B=R) and G
This 7 straight lines of=255, B=255, G=0, B=0, are divided into 3 regions：
According to the G of input, B values
If 1. point (G, B) is in region 1, then the mixing of new G values nG=R* (1-a)+B*a, R, B in proportion；
If 2. point (G, B) is in region 2, then new G values nG=R*a+B* (1-a), R, the mixing of B in proportion；
If 3. point (G, B) is in region 3, then new G values nG=G；It is constant.
Basic shade algorithm
The effect of basic shade is one shade of generation, and solid background is removed for scratching：For the image of green background, background
Middle g-b value is than larger, and g-b value is relatively small in prospect, and front and rear scape can be told with this.Introduce background sample point
(R, G, B), it is contemplated herein that sample point is green, i.e. G>R, G>B.Now there are following several situation explanations：
g-b<0：It is not green to show this pixel, then needs to retain, i.e. shade M need to take 255 (white)；
0<g-b<G-B：Show the degree of this pixel green not green between sample point, shade M need to 0~255 it
Between；
G-B<g-b：Show that this pixel is greener than sample point, then need to remove, i.e. shade M need to take 0 (black).
Therefore function representation is：
Similarly the situation to g-r is analyzed, and draws following formula：
For B axle, H is taken2=g, H2Above section, i.e. H2<B, g-b<0, i.e. M=are white；H1It is relevant with sample point, if taking two
Individual sample point highcolor, lowcolor, sample point G-B：Compare_a=ghigh-bhigh+glow-blow, is designated as G-
B；H1=g-k* (G-B), introduce parameter k, the deviation for correcting the sample point selected by user, H herein1Following part, i.e. b<
H1(k*(G-B)<G-b), i.e., M=is black；For R axles similarly, C is taken2=r, C1=g-k* (G-R).
For nine different zones, its value has nine kinds of situations.
Details shade algorithm
The effect of details shade is that, according to brightness of image and sample point brightness, one details shade of generation enters to basic shade
Row enhancing.Its implementation is：Input value highcolor (rgb), lowcolor (rgb), current pixel point (rgb)；According to bright
Spend formula：Brightness=0.29*r+0.59*g+0.12*b, obtains highcolor corresponding brightness luminhigh, lowcolor correspondence bright
Luminlow is spent, current pixel point brightness is lumin.
Brief description of the drawings
Fig. 1 overflows balance schematic diagram.
Fig. 2 overflows balanced algorithm and improves figure.
Fig. 3 overflows algorithm and improves figure again.
Fig. 4 bases shade algorithm pattern.
Fig. 5 details shade algorithm patterns.
Fig. 6 is flow chart of the invention.
Fig. 7 is Background.
Fig. 8 is scratched as result figure.
Fig. 9 synthetic effect figures.
Figure 10 foreground picture artworks.
Embodiment
Scratch as the workflow of part is：Pending video is inputted, then the chroma key for scratching picture is selected in preview region.
After this, system according to user's selection scratch as processing, that is, rejects background color.Scratch afterwards as result output, with background
Picture is merged (superposition)；Image after the completion of final output.
Perform step：
S1.CPU obtains related data, then that the foreground image data and background image data of rgb format is incoming from RAM
GPU global memories；
S2.CPU applies for space to GPU global memory, for being stored to the view data after synthesis；
The block of each in S3.GPU starts to perform Kernel functions；
S4. the data of a pixel in foreground image and background image are polymerize from global memory read it is shared
In internal memory；
S5. thread is synchronized, it is ensured that the data of processing all read in shared drive；
S6. foreground image and background image data are read out from shared drive；
S7. scratch as processing, result is written into the global memory of storage composograph；
S8. thread is synchronized, it is ensured that correlated results is all write in shared drive；
S9. the composograph data aggregate write-in global memory that will be stored in shared drive；
Result is transmitted back to CPU by S10.CPU from GPU；
S11.CPU shows the result of final image processing.
Claims (2)
1. a kind of stingy picture synthetic method based on CUDA, it is characterised in that：First, it is non-to foreground image using color balance operator
Green area carries out spilling Balance Treatment；Then a shade is generated using basic shade operator, scratches and remove solid background；Next
Basic shade is strengthened using details shade operator；After above-mentioned processing, final shade is formed, is carried out with background image
Fusion, the image after output synthesis；
It make use of GPU to accelerate in stingy picture and synthesis phase, further increase splicing speed；Test result indicates that, this method energy
Enough real-time image scratchings for effectively realizing video, frame per second reaches 10 frames, video keying can be more disclosure satisfy that compared to traditional CPU versions
Requirement of real time；
Overflow balanced algorithm：
Balanced algorithm is overflowed to for G in foreground picture>R point, there is nG=R, so reduces G values in the point of green, letter
Number is expressed as：
<mrow>
<mi>n</mi>
<mi>G</mi>
<mo>=</mo>
<mi>F</mi>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>,</mo>
<mi>R</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mi>R</mi>
</mtd>
<mtd>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>></mo>
<mi>R</mi>
<mo>)</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>G</mi>
</mtd>
<mtd>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>&le;</mo>
<mi>R</mi>
<mo>)</mo>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mrow>
<mi>n</mi>
<mi>G</mi>
<mo>=</mo>
<mi>F</mi>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>,</mo>
<mi>R</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mi>R</mi>
</mtd>
<mtd>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>></mo>
<mi>R</mi>
<mo>)</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>G</mi>
</mtd>
<mtd>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>&le;</mo>
<mi>R</mi>
<mo>)</mo>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
It is improved to overflowing balanced algorithm：For G<(B+R)/2, i.e. region 3, pixels illustrated point is not green, then is protected
Stay processing；For G>(B+R)/2, i.e. region 1 and 2, pixels illustrated point G values are too high, then are balanced processing, take new G=(R+
B)/2, it is changed into (R+B)/2 equivalent to R in (1) formula；Its function expression is：
<mrow>
<mi>n</mi>
<mi>G</mi>
<mo>=</mo>
<mi>F</mi>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>,</mo>
<mi>B</mi>
<mo>,</mo>
<mi>R</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mfrac>
<mrow>
<mi>R</mi>
<mo>+</mo>
<mi>B</mi>
</mrow>
<mn>2</mn>
</mfrac>
</mtd>
<mtd>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>></mo>
<mfrac>
<mrow>
<mi>R</mi>
<mo>+</mo>
<mi>B</mi>
</mrow>
<mn>2</mn>
</mfrac>
<mo>)</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>G</mi>
</mtd>
<mtd>
<mrow>
<mo>(</mo>
<mi>G</mi>
<mo>&le;</mo>
<mfrac>
<mrow>
<mi>R</mi>
<mo>+</mo>
<mi>B</mi>
</mrow>
<mn>2</mn>
</mfrac>
<mo>)</mo>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Next it is improved again, introduces parameter a, span is 0-1；
Using B=R straight lines as symmetry axis, parameter a determines ∠ ACD size, as a=1, ACD=0 ° of angle, as a=0, ∠ ACD
=90 °, i.e. ∠ ACD=(1-a) * 90, it can thus be appreciated that straight line AC slopes, obtain AC linear equations, you can distinguishable region 1；Parameter a phases
When in the value change and the size of reservation region that change F (G, B, R)；
Region 1：F (G, B, R)=a*R+ (1-a) * B；
Region 2：F (G, B, R)=(1-a) * R+a*B；
Region 3：F (G, B, R)=G；
It is identical with (2) formula as a=0.5, therefore function expression is changed into：
For convenience of being calculated, BR plans above are switched into BG plans；Abscissa represents G values (0-255), ordinate in way
B values (0-255) are represented, a tangent plane in RGB color is represented；
Input (R, G, B) is source images current pixel point, and a is the parameter spill for overflowing balance, can be according to the R values of input
Point C (R, R) is found in GB plane right-angle coordinate；Point A (R* (1-a), 0), point B ((255-R) * a+R, 255) are obtained according to a
A is ratio, makes A points can only be in 0-R scopes, B points can only be in the range of R-255；Two straight lines can be made by three above point
Line1：B=(G-R* (1-a))/a, Line2：B=(G-R)/a+R；According to Line1, Line2, Line3 (B=R) and G=
This 7 straight lines of 255, B=255, G=0, B=0, are divided into 3 regions：
According to the G of input, B values
If 1. point (G, B) is in region 1, then the mixing of new G values nG=R* (1-a)+B*a, R, B in proportion；
If 2. point (G, B) is in region 2, then new G values nG=R*a+B* (1-a), R, the mixing of B in proportion；
If 3. point (G, B) is in region 3, then new G values nG=G；It is constant；
Basic shade algorithm
The effect of basic shade is one shade of generation, and solid background is removed for scratching：For the image of green background, g- in background
B value is than larger, and g-b value is relatively small in prospect, and front and rear scape can be told with this；Introducing background sample point (R, G,
B), it is contemplated herein that sample point is green, i.e. G>R, G>B；Now there are following several situation explanations：
g-b<0：It is not green to show this pixel, then needs to retain, i.e. shade M need to take 255 (white)；
0<g-b<G-B：Show the degree of this pixel green not green between sample point, shade M need to be between 0~255；
G-B<g-b：Show that this pixel is greener than sample point, then need to remove, i.e. shade M need to take 0 (black)；
Therefore function representation is：
Similarly the situation to g-r is analyzed, and draws following formula：
For B axle, H is taken2=g, H2Above section, i.e. H2<B, g-b<0, i.e. M=are white；H1It is relevant with sample point, if taking two samples
Point highcolor, lowcolor, sample point G-B：Compare_a=ghigh-bhigh+glow-blow, is designated as G-B；H1=
G-k* (G-B), introduce parameter k, the deviation for correcting the sample point selected by user, H herein1Following part, i.e. b<H1(k*
(G-B)<G-b), i.e., M=is black；For R axles similarly, C is taken2=r, C1=g-k* (G-R)；
Details shade algorithm
The effect of details shade is that, according to brightness of image and sample point brightness, one details shade of generation increases to basic shade
By force；Its implementation is：Input value highcolor (rgb), lowcolor (rgb), current pixel point (rgb)；It is public according to brightness
Formula：Brightness=0.29*r+0.59*g+0.12*b, obtains highcolor corresponding brightness luminhigh, lowcolor corresponding brightness
Luminlow, current pixel point brightness is lumin.
2. a kind of stingy picture synthetic method based on CUDA according to claim 1, it is characterised in that：Scratch the work as part
Flow is：Pending video is inputted, then the chroma key for scratching picture is selected in preview region；After this, system is according to user
Selection scratch as processing, that is, rejects background color；Scratch afterwards as result output, carry out merging/being superimposed with background frame；Finally
Image after the completion of output；
Perform step：
S1.CPU obtains related data, then by the foreground image data and background image data of rgb format from the incoming GPU of RAM
Global memory；
S2.CPU applies for space to GPU global memory, for being stored to the view data after synthesis；
The block of each in S3.GPU starts to perform Kernel functions；
S4. the data of a pixel in foreground image and background image are polymerize from global memory and reads shared drive
In；
S5. thread is synchronized, it is ensured that the data of processing all read in shared drive；
S6. foreground image and background image data are read out from shared drive；
S7. scratch as processing, result is written into the global memory of storage composograph；
S8. thread is synchronized, it is ensured that correlated results is all write in shared drive；
S9. the composograph data aggregate write-in global memory that will be stored in shared drive；
Result is transmitted back to CPU by S10.CPU from GPU；
S11.CPU shows the result of final image processing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710336870.XA CN107194867A (en) | 2017-05-14 | 2017-05-14 | A kind of stingy picture synthetic method based on CUDA |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710336870.XA CN107194867A (en) | 2017-05-14 | 2017-05-14 | A kind of stingy picture synthetic method based on CUDA |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107194867A true CN107194867A (en) | 2017-09-22 |
Family
ID=59873255
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710336870.XA Pending CN107194867A (en) | 2017-05-14 | 2017-05-14 | A kind of stingy picture synthetic method based on CUDA |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107194867A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110335288A (en) * | 2018-09-26 | 2019-10-15 | 惠州学院 | A kind of video foreground target extraction method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101588459A (en) * | 2009-06-26 | 2009-11-25 | 北京交通大学 | A kind of video keying processing method |
| CN101882311A (en) * | 2010-06-08 | 2010-11-10 | 中国科学院自动化研究所 | Background modeling acceleration method based on CUDA (Compute Unified Device Architecture) technology |
| US20130064465A1 (en) * | 2011-09-12 | 2013-03-14 | Canon Kabushiki Kaisha | Image compression and decompression for image matting |
| CN103366364A (en) * | 2013-06-07 | 2013-10-23 | 太仓中科信息技术研究院 | Color difference-based image matting method |
| CN103581571A (en) * | 2013-11-22 | 2014-02-12 | 北京中科大洋科技发展股份有限公司 | Video image matting method based on three elements of color |
| US20140301639A1 (en) * | 2013-04-09 | 2014-10-09 | Thomson Licensing | Method and apparatus for determining an alpha value |
-
2017
- 2017-05-14 CN CN201710336870.XA patent/CN107194867A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101588459A (en) * | 2009-06-26 | 2009-11-25 | 北京交通大学 | A kind of video keying processing method |
| CN101882311A (en) * | 2010-06-08 | 2010-11-10 | 中国科学院自动化研究所 | Background modeling acceleration method based on CUDA (Compute Unified Device Architecture) technology |
| US20130064465A1 (en) * | 2011-09-12 | 2013-03-14 | Canon Kabushiki Kaisha | Image compression and decompression for image matting |
| US20140301639A1 (en) * | 2013-04-09 | 2014-10-09 | Thomson Licensing | Method and apparatus for determining an alpha value |
| CN103366364A (en) * | 2013-06-07 | 2013-10-23 | 太仓中科信息技术研究院 | Color difference-based image matting method |
| CN103581571A (en) * | 2013-11-22 | 2014-02-12 | 北京中科大洋科技发展股份有限公司 | Video image matting method based on three elements of color |
Non-Patent Citations (2)
| Title |
|---|
| 谭雄素等: "U-EDIT抠像技术在微课制作中的应用——以微课作品《重阳那些事》为例", 《数字教育》 * |
| 陈翔: "基于CUDA的抠像算法设计优化", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110335288A (en) * | 2018-09-26 | 2019-10-15 | 惠州学院 | A kind of video foreground target extraction method and device |
| CN110516534A (en) * | 2018-09-26 | 2019-11-29 | 惠州学院 | A kind of method for processing video frequency and device based on semantic analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sinha et al. | Image-based rendering for scenes with reflections | |
| US8351689B2 (en) | Apparatus and method for removing ink lines and segmentation of color regions of a 2-D image for converting 2-D images into stereoscopic 3-D images | |
| US7532752B2 (en) | Non-photorealistic sketching | |
| CN106485720A (en) | Image processing method and device | |
| Gao et al. | Detail preserved single image dehazing algorithm based on airlight refinement | |
| CN102779351A (en) | Interactive grayscale image colorizing method based on local linear model optimization | |
| US20140204125A1 (en) | Systems and methods for creating photo collages | |
| CN102098528A (en) | Method and device for converting planar image into stereoscopic image | |
| CN116997933A (en) | Method and system for constructing facial position map | |
| CN102750685A (en) | Image processing method and device | |
| CN110992247A (en) | Method and system for realizing special effect of straightening hair of portrait photo | |
| Grogan et al. | User interaction for image recolouring using£ 2 | |
| CN116563459A (en) | Text-driven immersive open scene neural rendering and mixing enhancement method | |
| Xiao et al. | Image hazing algorithm based on generative adversarial networks | |
| Seo et al. | Image recoloring using linear template mapping | |
| Zhang et al. | Refilming with depth-inferred videos | |
| Liu | Two decades of colorization and decolorization for images and videos | |
| Chang et al. | A self-adaptive single underwater image restoration algorithm for improving graphic quality | |
| CN107194867A (en) | A kind of stingy picture synthetic method based on CUDA | |
| CN112132923A (en) | Two-stage digital image style transformation method and system based on style thumbnail high-definition | |
| Ye et al. | Hybrid scheme of image’s regional colorization using mask r-cnn and Poisson editing | |
| Liao et al. | Depth annotations: Designing depth of a single image for depth-based effects | |
| CN110689609B (en) | Image processing method, image processing device, electronic equipment and storage medium | |
| CN117011324A (en) | Image processing method, device, electronic equipment and storage medium | |
| CN115170921A (en) | Binocular stereo matching method based on bilateral grid learning and edge loss |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170922 |