CN101764915A - Key evolutionary method and device based on bounding boxes - Google Patents
Key evolutionary method and device based on bounding boxes Download PDFInfo
- Publication number
- CN101764915A CN101764915A CN200810239437A CN200810239437A CN101764915A CN 101764915 A CN101764915 A CN 101764915A CN 200810239437 A CN200810239437 A CN 200810239437A CN 200810239437 A CN200810239437 A CN 200810239437A CN 101764915 A CN101764915 A CN 101764915A
- Authority
- CN
- China
- Prior art keywords
- key
- matrix
- pixel
- bounding box
- patch
- 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
Images
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to a key evolutionary method and a device based on bounding boxes. The method comprises the following steps: establishing key patches by using pixel points of key images processed by a warp method selected by users; calculating the bounding boxes in the RGB three-dimensional space of each key patch according to the RGB value of the given pixel; calculating a converting matrix M according to the RGB value of the given pixel; converting the bounding boxes of the key patches corresponding to the matrix M; and using the key patches for key value regulation. Each embodiment of the invention can solve the problem that the image processed by the key method in the prior art is not ideal enough. The results (i.e. the key images) of some key methods are modified, so the goal of key evolutionary effect is reached.
Description
Technical field
The invention belongs to field of video image processing, relate in particular to the special effect processing and synthetic a kind of key tuning method and the device that are used for video image based on bounding box.
Background technology
In the image/video process software, it is a kind of demand very widely that a certain class color in the image/video is scratched.The given a kind of color of the essence of these class methods (also may add other parameter) all calculates a key assignments to each pixel in the image, the neutralize key assignments of the identical or approaching color pixel of this color of image is zero or smaller, and it is bigger or be 1 that color differs the key assignments of bigger pixel.This key assignments just can be used as this image and the background image alpha passage when synthesizing so, thereby reaches the purpose of scratching picture.
The key method has a lot, and chroma key is arranged, and the brightness key is directly based on key of rgb space or the like.These methods respectively have individual merits and demerits.But generally speaking, under the more scattered situation of the distribution of color of the image that will scratch, all be difficult to draw satisfied effect with the sort of key algorithm.
Summary of the invention
At the defective that exists in the prior art, the purpose of this invention is to provide a kind of key tuning method and device thereof based on bounding box.This method and system can be adjusted to customer satisfaction system image effect with original key image by high quality and high efficiency.
For finishing the foregoing invention purpose, the invention provides a kind of key tuning method based on bounding box, may further comprise the steps:
(1) pixel that utilizes the user to choose is created the key patch;
(2), calculate the Rectangular Bounding Volume of each key patch in RGB three dimensions according to the rgb value of given pixel;
(3), calculate a transform matrix M according to the rgb value of given pixel;
(4) bounding box with the key patch of transform matrix M correspondence carries out conversion;
(5) the utility key patch carries out the key assignments adjustment.
Further, the set of the pixel chosen for the user of described key patch.
Further, the method for calculating bounding box in the described step (2) comprises: create a pixel S set for each key patch, all pixels among the pair set S are at R-G, G-B, and fitting a straight line is carried out in the projection of B-R, obtains translation matrix Mt
0With spin matrix Mr
0All pixels among the pair set S are used Mr again
0* Mt
0Transformation matrix is obtained spin matrix Mr
1At last with the application matrix Mr again of all pixels in the S set
1, obtain pixel after all conversion at the x axial coordinate, the maximum and the minimum value of y axial coordinate and z axial coordinate constitute a bounding box.
Further, calculating transformation matrix in the described step (3) further may further comprise the steps:
1) create a pixel S set for each key patch;
2) all pixels among the pair set S are at R-G, G-B, and fitting a straight line is carried out in the projection of B-R, obtains translation matrix Mt
0With spin matrix Mr
0
3) all pixels among the pair set S are used Mr
0* Mt
0Transformation matrix is obtained spin matrix Mr
1
4) all pixels among the pair set S are used Mr
1* Mr
0* Mt
0Transformation matrix is obtained translation matrix Mt
1With scaled matrix Ms;
5) all pixels among the pair set S are used M=Ms*Mt
1* Mr
1* Mr
0* Mt
0
Further, in the described step (4) bounding box of the key patch of transform matrix M correspondence being carried out conversion is that bounding box is transformed to a central point at initial point, and the length of side is in 2 the cube;
Further, the utility key patch further may further comprise the steps in the described step (4):
1) calculates soft_threshold and soft_span according to softness;
2) use transform matrix M for the rgb value of each pixel in the image, obtain new pixel r ', g ', b ';
3) calculate the bounding box weight w of each pixel;
4) weight w, soft_threshold and the soft_span according to each pixel calculates interpolation factor f;
5) between the former key assignments of pixel and purpose value, be that interpolation factor carries out linear interpolation with f.
Further, the equation of described fitting a straight line is:
y＝s·x+c
According to least square method, we can get:
c＝(∑y·∑x
2-∑x·∑(x·y)/(N·∑x
2-(∑x)
2)
s＝(N·∑(x·y)-∑x·∑y)/(N·∑x
2-(∑x)
2)
In the following formula ∑ x be meant all sampled point x values and; ∑ y be meant all sampled point y values and; ∑ x
2Be meant all sampled point x values square and; ∑ (xy) be meant all sampled point x values and y machine on duty and; N is the sampled point number.
For finishing the foregoing invention purpose, the present invention also provides a kind of key tuning device based on bounding box, comprises with lower module: key patch extraction module, bounding box computing module, transformation matrix and bounding box conversion module;
Described key patch extraction module receives the key image of input, and the pixel that the user is chosen constitutes different key patches and sends into bounding box computing module and bounding box conversion module.
Described bounding box computing module is created a pixel set for each the key patch that receives, and all pixels are obtained translation matrix Mt in the pair set
0With spin matrix Mr
0And Mr
1, use spin matrix Mr
1, obtain pixel after all conversion at the x axial coordinate, the maximum and the minimum value of y axial coordinate and z axial coordinate constitute a bounding box, and deliver to transformation matrix.
Described transformation matrix utilizes the bounding box that receives to make up a translation matrix Mt
1With a scaled matrix Ms, obtain final transform matrix M=Ms*Mt
1* Mr
1* Mr
0* Mt
0After deliver to key patch conversion module.
Described bounding box conversion module is used transform matrix M to RGB three dimensions each pixel of input picture key patch and the relation of bounding box is adjusted original key assignments, and output is through adjusted image.
Further, described key patch extraction module receives the key image of input, and the pixel of the key image that the user is chosen can be choosing of a pixel of a pixel, also can be the frame choosing.
Further, described bounding box conversion module is that calculating realizes through the rgb value and the described cubical relation of the pixel of transform matrix M to original key assignments adjustment.
The present invention has tangible advantage and good effect.Utilize key tuning method and the device based on bounding box of the present invention, ask the transformation matrix of key patch bounding box to separate fully with this key patch of application, in an image sequence, transformation matrix only need be asked once, the then every frame of the method for utility key patch is carried out one time, and method is simple, and optimizes easily, therefore whole efficient is very high, makes adjusted image can reach the fine effect that gets.
Description of drawings
Fig. 1 is the implementing procedure figure of institute of the present invention method;
Fig. 2 is the method implementing procedure figure that matrix is changed in the changes persuing of the method for the invention;
Fig. 3 is the implementing procedure figure of the utility key patch of the method for the invention;
Fig. 4 is the structure chart of device of the present invention;
Fig. 5 is a video image design sketch of the present invention;
Fig. 6 is according to the key image schematic diagram after the stingy background blueness of the present invention;
Fig. 7 is according to the key image schematic diagram after the application black keys patch of the present invention.
Embodiment
Below in conjunction with the drawings and specific embodiments the present invention is further described.
Fig. 1 is the implementing procedure figure of key tuning method of the present invention, and as shown in Figure 1, a kind of key tuning method based on bounding box may further comprise the steps:
At first at step S101, the pixel that the user chooses is created a key patch, the set of the pixel that this key patch (key patch) is chosen for the user, scratching picture at video image, to handle Central Plains should be that the key patch of all scratching is referred to as black key patch (black key patch), and former should be that the key patches that all keep are referred to as white key patch (black key patch);
At step S102,, calculate the bounding box of each key patch in RGB three dimensions according to the rgb value of given pixel.The bounding box computational methods are as follows: create the S set of a pixel at first for each key patch; Secondly all pixels among pair set S are at R-G, G-B, and fitting a straight line is carried out in the projection of B-R, obtains translation matrix Mt
0With spin matrix Mr
0All pixels among the pair set S are used Mr once more
0* Mt
0Transformation matrix is obtained spin matrix Mr
1At last with the application matrix Mr again of all pixels in the S set
1, obtain pixel after all conversion at the x axial coordinate, the maximum and the minimum value of y axial coordinate and z axial coordinate, these maximums and minimum value form a Rectangular Bounding Volume.
At step S103,, calculate a transform matrix M according to the rgb value of given pixel;
At step S104, the bounding box of the key patch of transform matrix M correspondence is carried out conversion, just the bounding box with the pairing key patch of transform matrix M transforms to a central point at initial point, and the length of side is in 2 the cube;
At step S105, the utility key patch carries out the key assignments adjustment.Carry out interpolation with interpolation factor at original key assignments of pixel and target key assignments (if black key patch target key assignments is exactly 0, if white key patch target key assignments is exactly 1), the result of interpolation is adjusted key assignments, and the pixel adjustment that the user chooses is finished.
Fig. 2 is the algorithm implementing procedure figure that matrix is changed in the changes persuing of the method for the invention, and with reference to figure 2, the algorithm implementing procedure that matrix is changed in changes persuing of the present invention may further comprise the steps:
At first, create the S set of a pixel for each key patch at step S201;
At step S202, all pixels among the pair set S are at R-G, G-B, and fitting a straight line is carried out in the projection of B-R, obtains translation matrix Mt
0With spin matrix Mr
0
The rgb value of the middle pixel of certain key patch is projected to R-G, G-B, B-R gets on three planes, carries out the fitting a straight line of two dimension then on each plane, and in fact we only are concerned about the direction vector of fitting a straight line.Fitting method is as follows:
We suppose that the equation of fitting a straight line is:
y＝s·x+c
According to least square method, we can get:
c＝(∑y·∑x
2-∑x·∑(x·y)/(N·∑x
2-(∑x)
2)
s＝(N·∑(x·y)-∑x·∑y)/(N·∑x
2-(∑x)
2)
In the following formula ∑ x be meant all sampled point x values and; ∑ y be meant all sampled point y values and; ∑ x
2Be meant all sampled point x values square and; ∑ (xy) be meant all sampled point x values and y machine on duty and; N is the sampled point number.
We just can calculate the direction vector of this straight line after calculating c and s.Attention: this method is invalid (N ∑ x to the fitting a straight line perpendicular to the x axle
2-(∑ x)
2Be 0), so we need be rotated counterclockwise the x-y plane fitting a straight line that tries again after 90 degree, the direction vector of the straight line after match dextrorotation again turn 90 degrees as final direction vector.But in the time of mostly, the fitting a straight line of acquiescence all is effectively with revolving the fitting a straight line that turn 90 degrees, and need compare employed variance in the least square method this time so, sees which is little just with which direction vector as fitting a straight line.
The formula of the variance of straight line least square method is: ∑ (y-sx-c)
2
After the expansion be:
∑y
2+s
2·∑x
2+N·c
2-2·s·∑(x·y)-2·c·∑y+2·s·c·∑x
Carrying out fitting a straight line after this step, we have obtained three direction vectors
Corresponding R-G respectively, G-B, the direction vector of the fitting a straight line on three planes of B-R, and then calculate these three vectors are being rotated counterclockwise 90 degree in plane separately vectors
We ask cross product in certain sequence successively in twos to these three new vectors then, and that vector that takes out mould value maximum among these three cross product results is designated as
Use then
Again and
Carry out cross product successively, that vector that takes out mould value maximum among the cross product result is designated as
Calculate at last
Like this,
The direction that is exactly three quadratures is an amount of, and we can construct a rotation transformation matrix M r by the knowledge of geometric transformation in the computer graphics
0, this rotation transformation matrix can be with above the former reference axis
Three transform vectors go on three unit vectors of reference axis.
Like this, we are at first translations of point all in the key patch, make after the translation that initial point has been arrived at the center of all points just in the key patch, and we remember that this translation matrix is Mt
0
In step 203, all pixels among the pair set S are used Mr
0* Mt
0Transformation matrix is obtained spin matrix Mr
1All pixels among the pair set S are used the rotation transformation matrix M r that calculates above
0With translation matrix be Mt
0, application matrix Mr
0* Mt
0Then these spot projections after the conversion are gone to the x-y plane, use front structural matrix Mr again
0Method is carried out the fitting a straight line on x-y plane, obtains a spin matrix Mr
1
In step 204, all pixels among the pair set S are used Mr
1* Mr
0* Mt
0Transformation matrix is obtained translation matrix Mt
1With scaled matrix Ms.All pixels in the S set are used this matrix M r again
1, obtain pixel after all conversion at the x axial coordinate, the maximum and the minimum value of y axial coordinate and z axial coordinate are with front structural matrix Mr
0Method construct goes out a translation matrix Mt
1Pixel after all conversion is at the x axial coordinate, and the maximum of y axial coordinate and z axial coordinate and minimum value form a bounding box.The central point of this bounding box is moved to initial point, use front structural matrix Mr then
0Method is constructed a scaled matrix Ms again, makes bounding box all zoom on (1,1) in each reference axis.
In step 205, all pixels among the pair set S are used Ms*Mt
1* Mr
1* Mr
0* Mt
0Be final transformation matrix.
Fig. 3 is the implementing procedure figure of the utility key patch of the method for the invention, and with reference to figure 1, the utility key patch of the method for the invention may further comprise the steps:
At first, calculate soft_threshold and soft_span according to softness at step S301.Each key patch all has two extra parameters, and one is range, and one is that (soft limit degree, it can control the soft degree that is adjusted pixel region and is not adjusted border between the pixel region to softness.), they are used for the degree that the operating key patch is used, and in order to play control action, at first introduce two new variable soft_threshold and soft_span.In the time of softness＜=0.1, soft_threshold=1, soft_span=softness; In the time of softness＞0.1,
soft_threshold＝1-2*(softness-0.1)/0.9，soft_span＝1.1-soft_threshold；
In step 302, use transform matrix M for the rgb value of each pixel in the image, obtain new pixel r ', g ', b ';
In step 303, calculate the bounding box weight w of each pixel., this value calculating method is as follows:
If the RGB homogeneous coordinates column vector form of pixel is
w＝max(|r’|，|g’|，|b’|)/range
In step 304, calculate interpolation factor f according to weight w, soft_threshold and the soft_span of each pixel, the computational methods of interpolation factor f are as follows:
f＝0
If w＜=soft_threshold
f＝min(1，(w-soft_threshold)/soft_span)
If w＞soft_threshold
In step 305, between the former key assignments of pixel and purpose value (the black key patch is 0, and white key patch is 1), be that interpolation factor carries out linear interpolation with f, the result of interpolation is adjusted key assignments.
In the present embodiment, method of the present invention is to make amendment on result's (being the key image) of certain strong method, thereby reaches the purpose of optimizing the key effect.The unsatisfied place of result's (being the key image) after the user handles through kind of strong method piece image, key assignments such as some pixel that should scratch fully is still very big, or some key values of pixels that should keep fully is smaller, at the unfavorable pixel of these effects, the user can by the input image on selected pixels method (can be a pixel of a pixel choose, also can be frame choosing) sign comes out, and tell these pixels of method should be all scratch or should be all to keep.The set of the pixel that these users choose is a key patch (key patch), former should be that the key patch of all scratching is referred to as black key patch (black key patch), and former should be that the key patches that all keep are referred to as white key patch (black key patch).These key patches have been arranged, this method calculates the bounding box of a rectangle can for each key patch in the three dimensions of RGB, according to the relation of each pixel of input picture and bounding box original key assignments is adjusted (such as given black key a patch then, the key assignments that is positioned at the point at bounding box center will become 0 certainly, weak more to export-oriented 0 trend that changes more), thus reach the purpose of key tuning.
As shown in Figure 4, a kind of key tuning device based on bounding box comprises with lower module: key patch extraction module, bounding box computing module, transformation matrix and bounding box conversion module.
Key patch extraction module receives through the key image after the processing of key method, and the unsatisfied pixel of treatment effect that the user is chosen constitutes white key patch and black key patch respectively, delivers to bounding box computing module and bounding box conversion module.
The bounding box computing module is created each the key patch that receives the S set of a pixel; And all pixels among the pair set S are at R-G, G-B, and fitting a straight line is carried out in the projection of B-R, obtains translation matrix Mt
0With spin matrix Mr
0All pixels among the pair set S are used Mr again
0* Mt
0Transformation matrix is obtained spin matrix Mr
1At last with the application matrix Mr again of all pixels in the S set
1, obtain pixel after all conversion at the x axial coordinate, the maximum and the minimum value of y axial coordinate and z axial coordinate constitute a bounding box, and deliver to transformation matrix.
Transformation matrix utilizes the bounding box that receives to make up a translation matrix Mt
1, the central point of bounding box moved to initial point after, construct a scaled matrix Ms again, make bounding box all zoom on (1,1) in each reference axis, obtain final transform matrix M=Ms*Mt
1* Mr
1* Mr
0* Mt
0After deliver to key patch conversion module.
The bounding box conversion module is used transform matrix M to RGB three dimensions, and the bounding box of the pairing key patch of M is transformed to a central point at initial point, and the length of side is in 2 the cube, the key assignments of key patch to be adjusted, and obtains adjusted key image.
In the present embodiment, utilize key tuning method and the device based on bounding box of the present invention, ask the transformation matrix of key patch bounding box to separate fully with this key patch of application, in an image sequence, transformation matrix only need be asked once, and the then every frame of the method for utility key patch is carried out one time, method is comparatively simple, and than being easier to optimization, therefore whole efficient is very high, makes adjusted image can reach the fine effect that gets.
Invent described method and apparatus and be not limited to the embodiment described in the embodiment, those skilled in the art's technical scheme according to the present invention draws other execution mode, belongs to technological innovation scope of the present invention equally.
Claims (10)
1. key tuning method based on bounding box, this method may further comprise the steps:
(1) pixel that utilizes the user to choose is created the key patch;
(2), calculate the Rectangular Bounding Volume of each key patch in RGB three dimensions according to the rgb value of given pixel;
(3), calculate a transform matrix M according to the rgb value of given pixel;
(4) bounding box with the key patch of transform matrix M correspondence carries out conversion;
(5) the utility key patch carries out the key assignments adjustment.
2. method according to claim 1 is characterized in that: the set of the pixel that described key patch is chosen for the user.
3. method according to claim 1, it is characterized in that: the method for calculating bounding box in the described step (2) comprises: create a pixel S set for each key patch, all pixels among the pair set S are at R-G, G-B, fitting a straight line is carried out in the projection of B-R, obtains translation matrix Mt
0With spin matrix Mr
0All pixels among the pair set S are used Mr again
0* Mt
0Transformation matrix is obtained spin matrix Mr
1At last with the application matrix Mr again of all pixels in the S set
1, obtain pixel after all conversion at the x axial coordinate, the maximum and the minimum value of y axial coordinate and z axial coordinate constitute a bounding box.
4. method according to claim 1 is characterized in that: calculate transform matrix M in the described step (3) and further may further comprise the steps:
1) create a pixel S set for each key patch;
2) all pixels among the pair set S are at R-G, G-B, and fitting a straight line is carried out in the projection of B-R, obtains translation matrix Mt
0With spin matrix Mr
0
3) all pixels among the pair set S are used Mr
0* Mt
0Transformation matrix is obtained spin matrix Mr
1
4) all pixels among the pair set S are used Mr
1* Mr
0* Mt
0Transformation matrix is obtained translation matrix Mt
1With scaled matrix Ms;
5) all pixels among the pair set S are used M=Ms*Mt
1* Mr
1* Mr
0* Mt
0
5. according to the described method of one of claim 1 to 4, it is characterized in that: in the described step (4) bounding box of the key patch of transform matrix M correspondence being carried out conversion is that bounding box is transformed to a central point at initial point, and the length of side is in 2 the cube;
6. method according to claim 5 is characterized in that: the utility key patch further may further comprise the steps in the described step (4):
1) calculates soft_threshold and soft_span according to softness;
2) use transform matrix M for the rgb value of each pixel in the image, obtain new pixel r ', g ', b ';
3) calculate the bounding box weight w of each pixel;
4) weight w, soft_threshold and the soft_span according to each pixel calculates interpolation factor f;
5) between the former key assignments of pixel and purpose value, be that interpolation factor carries out linear interpolation with f.
7. method according to claim 3 is characterized in that: the equation of described fitting a straight line is: y=sx+c
Wherein,
c＝(∑y·∑x
2-∑x·∑(x·y)/(N·∑x
2-(∑x)
2)
s＝(N·∑(x·y)-∑x·∑y)/(N·∑x
2-(∑x)
2)
In the following formula ∑ x be meant all sampled point x values and; ∑ y be meant all sampled point y values and; ∑ x
2Be meant all sampled point x values square and; ∑ (xy) be meant all sampled point x values and y machine on duty and; N is the sampled point number.
8. the key tuning device based on bounding box comprises with lower module: key patch extraction module, bounding box computing module, transformation matrix and bounding box conversion module;
Described key patch extraction module receives the key image of input, and the key image slices vegetarian refreshments that the user is chosen constitutes different key patches and sends into bounding box computing module and bounding box conversion module;
Described bounding box computing module is created the set of a pixel for each the key patch that receives, and all pixels are obtained translation matrix Mt in the pair set
0With spin matrix Mr
0And Mr
1, use spin matrix Mr
1, obtain pixel after all conversion at the x axial coordinate, the maximum and the minimum value of y axial coordinate and z axial coordinate constitute a Rectangular Bounding Volume, and deliver to transformation matrix;
Described transformation matrix utilizes the bounding box that receives to make up a translation matrix Mt
1With a scaled matrix Ms, obtain final transform matrix M=Ms*Mt
1* Mr
1* Mr
0* Mt
0After deliver to the bounding box conversion module;
Described bounding box conversion module is used transform matrix M to RGB three dimensions, and the bounding box of the pairing key patch of M is transformed to a central point at initial point, and the length of side is in 2 the cube, original key assignments to be adjusted, and output is through adjusted image.
9. a kind of key tuning device according to claim 8 based on bounding box, it is characterized in that: described key patch extraction module receives the key image of input, the pixel of the key image that the user is chosen, can be choosing of a pixel of a pixel, also can be the frame choosing.
10. it is characterized in that according to Claim 8 or 9 described a kind of key tuning devices based on bounding box: described bounding box conversion module is what to be calculated through the rgb value of the pixel of transform matrix M and described cubical relation realization to original key assignments adjustment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810239437A CN101764915A (en) | 2008-12-10 | 2008-12-10 | Key evolutionary method and device based on bounding boxes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810239437A CN101764915A (en) | 2008-12-10 | 2008-12-10 | Key evolutionary method and device based on bounding boxes |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101764915A true CN101764915A (en) | 2010-06-30 |
Family
ID=42495890
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810239437A Pending CN101764915A (en) | 2008-12-10 | 2008-12-10 | Key evolutionary method and device based on bounding boxes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101764915A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102456220A (en) * | 2010-10-25 | 2012-05-16 | 新奥特（北京）视频技术有限公司 | Color image noise channel extraction method based on bounding box |
| CN108876710A (en) * | 2018-06-20 | 2018-11-23 | 四川斐讯信息技术有限公司 | A kind of picture transform method and system |
| CN108986159A (en) * | 2018-04-25 | 2018-12-11 | 浙江森马服饰股份有限公司 | A kind of method and apparatus that three-dimensional (3 D) manikin is rebuild and measured |
-
2008
- 2008-12-10 CN CN200810239437A patent/CN101764915A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102456220A (en) * | 2010-10-25 | 2012-05-16 | 新奥特（北京）视频技术有限公司 | Color image noise channel extraction method based on bounding box |
| CN102456220B (en) * | 2010-10-25 | 2014-03-19 | 新奥特（北京）视频技术有限公司 | Color image noise channel extraction method based on bounding box |
| CN108986159A (en) * | 2018-04-25 | 2018-12-11 | 浙江森马服饰股份有限公司 | A kind of method and apparatus that three-dimensional (3 D) manikin is rebuild and measured |
| CN108986159B (en) * | 2018-04-25 | 2021-10-22 | 浙江森马服饰股份有限公司 | Method and equipment for reconstructing and measuring three-dimensional human body model |
| CN108876710A (en) * | 2018-06-20 | 2018-11-23 | 四川斐讯信息技术有限公司 | A kind of picture transform method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Multimodal transfer: A hierarchical deep convolutional neural network for fast artistic style transfer | |
| Xiao et al. | Brightness and contrast controllable image enhancement based on histogram specification | |
| Dong et al. | Color-guided depth recovery via joint local structural and nonlocal low-rank regularization | |
| CN111768477B (en) | Three-dimensional facial expression base establishment method and device, storage medium and electronic equipment | |
| CN110570377A (en) | group normalization-based rapid image style migration method | |
| CN111951384B (en) | Three-dimensional face reconstruction method and system based on single face picture | |
| Shao et al. | QoE-guided warping for stereoscopic image retargeting | |
| CN106652015B (en) | Virtual character head portrait generation method and device | |
| CN104506872B (en) | A kind of method and device of converting plane video into stereoscopic video | |
| CN101764915A (en) | Key evolutionary method and device based on bounding boxes | |
| CN108596992B (en) | Rapid real-time lip gloss makeup method | |
| CN113793255A (en) | Method, apparatus, device, storage medium and program product for image processing | |
| Lei et al. | GALFusion: Multi-exposure image fusion via a global–local aggregation learning network | |
| Lei et al. | Low-light image enhancement using the cell vibration model | |
| Duong et al. | Dmt-net: deep multiple networks for low-light image enhancement based on retinex model | |
| Wang et al. | Learning continuous depth representation via geometric spatial aggregator | |
| Yan et al. | Shape-optimizing mesh warping method for stereoscopic panorama stitching | |
| Hongying et al. | Image completion by a fast and adaptive exemplar-based image inpainting | |
| Liu et al. | Structure-preserving image completion with multi-level dynamic patches | |
| CN111669563A (en) | Stereo image visual comfort enhancement method based on reinforcement learning | |
| CN106408499B (en) | Method and device for acquiring reverse mapping table for image processing | |
| CN102456220B (en) | Color image noise channel extraction method based on bounding box | |
| Pan et al. | Salient structural elements based texture synthesis | |
| Wang et al. | A natural shape-preserving stereoscopic image stitching | |
| Zheng et al. | Efficient color-to-gray conversion for digital images in gradient domain |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20100630 |