CN102521807B - Method for transferring colors by utilizing color space distribution - Google Patents
Method for transferring colors by utilizing color space distribution Download PDFInfo
- Publication number
- CN102521807B CN102521807B CN 201110396172 CN201110396172A CN102521807B CN 102521807 B CN102521807 B CN 102521807B CN 201110396172 CN201110396172 CN 201110396172 CN 201110396172 A CN201110396172 A CN 201110396172A CN 102521807 B CN102521807 B CN 102521807B
- Authority
- CN
- China
- Prior art keywords
- color
- dominant color
- source images
- zone
- dominant
- 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
- 238000000034 method Methods 0.000 title claims abstract description 51
- 239000003086 colorant Substances 0.000 title abstract description 5
- 238000013507 mapping Methods 0.000 claims abstract description 42
- 238000004040 coloring Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000000638 solvent extraction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 230000003760 hair shine Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- VYMDGNCVAMGZFE-UHFFFAOYSA-N phenylbutazonum Chemical compound O=C1C(CCCC)C(=O)N(C=2C=CC=CC=2)N1C1=CC=CC=C1 VYMDGNCVAMGZFE-UHFFFAOYSA-N 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000002834 transmittance Methods 0.000 description 1
- 230000017260 vegetative to reproductive phase transition of meristem Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method for transferring colors by utilizing color space distribution. The method comprises the following steps of: by combination of color and space position information of image pixels, segmenting a source image and a target image into a plurality of dominant color areas; carrying out calculation on each dominant color area in the source image and the target image and obtaining the characteristic of each dominant color area; using a color mean and geometric center information of the dominant color areas to establish an optimal many-to-many probability map between the dominant color areas of the source image and the dominant color areas of the target image and leading the sum of the mapping cost among the dominant color areas to be minimum; and according to the optimal probability map, carrying out calculation on the dominant color areas of the source image again to obtain a new color mean and a new color variance, and updating the color of each pixel in the dominant color areas of the source image.
Description
Technical field
The invention belongs to the digital image processing techniques field, a kind of target image that utilizes that relates to computer graphics and Digital Image Processing is edited image as a reference, particularly the image processing method that picture tone is edited and strengthened.
Background technology
Along with a large amount of uses of image acquisition equipments such as digital camera, digital picture becomes very general media information.Because odjective causes such as capture environment, camera parameters, the original image that obtains does not often satisfy people's expectation, therefore need carry out post-processed by the graph and image processing technology.The image manipulation that some are basic as denoising, convergent-divergent, cutting, adjustment luminance contrast etc., can be handled by the image processing software (Photoshop etc.) of specialty.Yet when utilizing conventional image processing software to adjust the color harmony Color Style of image, the layman often is difficult to reach the effect of wanting.
The color of image transmission is a kind of picture editting's technology easily and effectively.Reinhard (REINHARD, E., ASHIKHMIN, M., GOOCH, B., AND SHIRLEY, P.2001.Color transfer between images.IEEE Comput.Graph.Appl.21 (September) 34-41.) has at first proposed global image color transmission method.This method can be delivered to the dominant color style of piece image in another width of cloth image, thereby reaches the purpose of automatic adjustment picture tone.The computing of global color transmission method is simple, but only just can obtain effect preferably when source images and target image have similar color distribution.Dong (DONG, W., BAO, G., ZHANG, X., AND PAUL, J.-C.2010.Fast local color transfer via dominant colors mapping.In ACM SIGGRAPH ASIA 2010Sketches, ACM, New York, NY, USA, SA ' 10,46:1-46:2.) utilize the method for extracting dominant color, in source images and target image, extract the dominant color of equal number, and between source images and target image dominant color, make up the mapping one to one of an optimum.When source images and target image had the dominant color of equal number, this method can guarantee that all dominant color of target image all pass to source images.If but the dominant color quantity of source images and target image is when inconsistent, the method for Dong can not obtain a gratifying result usually.
Goal of the invention
Being directed to source images dominant color quantity is less than under the situation of target image dominant color quantity, the dominant color of target image can not pass to this problem of source images fully, the invention provides all dominant color that can guarantee target image and all be delivered in the source images, and these target image dominant color can also keep a kind of method of utilizing color space to distribute color is transmitted of its original locus distribution in source images.
For achieving the above object, a kind of method of utilizing color space to distribute color is transmitted provided by the invention comprises that step is as follows:
Step S1: combining image color of pixel and spatial positional information, utilize the average drifting algorithm that source images and target image are divided into a plurality of dominant color zone;
Step S2: regard the pixel in the dominant color zone as a sample set, each dominant color zone in source images, the target image is calculated, obtain the following feature in each dominant color zone: all pixels that color average, color variance, geometric center and this dominant color zone comprise;
Step S3: color average and the geometric center information of utilizing the dominant color zone, between source images dominant color zone and target image dominant color zone, set up the probability mapping of a multi-to-multi, each dominant color zone of source images is mapped to each dominant color zone of target image with different probability, asks for source images dominant color zone and shine upon the mapping cost summation minimum that makes between the dominant color zone to probability optimum between the target image dominant color zone;
Step S4: according to the probability mapping of optimum, again source images dominant color zone is calculated, obtain new color average and the color variance in source images dominant color zone;
Step S5: utilize new color average and color variance, each color of pixel in the source images dominant color zone is upgraded.
Beneficial effect of the present invention: the present invention utilizes average drifting technology (Mean Shift) that image is cut apart, and obtains the dominant color zone of image; The probability that adopts earthquake distance method (Earth Mover ' s Distance) to set up a multi-to-multi then between source images dominant color zone and target image dominant color zone shines upon, and allows each dominant color zone is mapped in each target dominant color zone with the most appropriate probability in the source images; The dominant color of target image distributes because cut apart and set up the locus of all having considered color in two processes of mapping at image, so will be delivered in the source images together with its locus simultaneously; Even the dominant color quantity of source images and target image is inconsistent, can guarantee that also colors all in the target image all is applied to source images; Whole process of the present invention is automatic fully, and also can obtain result preferably under target image and source images dominant color quantity differ bigger situation.This method can not only pass to source images with the tone style of target image, and the locus that can also make the result images that obtains keep color of object distributes.
Description of drawings
Fig. 1 is the frame diagram of the inventive method;
Fig. 2 is algorithm workflow synoptic diagram;
Fig. 3 is dominant color zone mapping synoptic diagram;
Fig. 4 is dominant color region quantity inconsistent results figure;
The color transmission that Fig. 5 is based on the part semanteme is figure as a result.
Embodiment
Describe each related detailed problem in the technical solution of the present invention in detail below in conjunction with accompanying drawing.Be to be noted that described embodiment only is intended to be convenient to the understanding of the present invention, and it is not played any restriction effect.
As depicted in figs. 1 and 2, method of the present invention mainly is divided into three steps: image is cut apart; The mapping of dominant color zone; The source images color is upgraded.Adopt average drifting (Mean Shift) image partition method that image is cut apart, make adjacent and have similar color pixel and make as a whole participation color map; Between all source images dominant color zones and target image dominant color zone, set up the probability mapping of a multi-to-multi, the cost summation minimum that color is transmitted, thus each the dominant color zone in the source images all is mapped to several the most similar to it in target image dominant color zones; For some dominant color zone, utilize this dominant color zone and upgrade all color of pixel in this dominant color zone simultaneously in abutting connection with the mapping result in dominant color zone, thus the border between the smooth region.Described dominant color zone is the contiguous pixels zone that has similar color attribute in source images or the target image.
At describing in detail below the specific algorithm of each step.
For the locus that keeps color distributes, need when split image, consider the space continuity of color.Utilize Mean Shift partitioning algorithm (COMANICIU in the method for the present invention, D., AND MEER, P.2002.Mean shift:A robust approach toward feature space analysis.IEEE Trans.Pattern Anal.Mach.Intell.24 (May)) source images and target image are cut apart.Mean Shift algorithm is combined into the vector of one five dimension (coloured image) or three-dimensional (gray level image) with color of pixel and spatial positional information, and utilizes the non-parametric estmation method that the probability density of vector is estimated.Each extreme point of probability density all is considered as a cluster centre.Image can be divided into several zones according to these cluster centres, be referred to as the dominant color zone.To each dominant color zone, calculate the set that its all pixels of color average, color variance, geometric center coordinate and intra-zone are formed, and these information are used for representing this dominant color zone as the feature in dominant color zone.
The feature of calculating source images dominant color zone comprises as follows:
In the formula, subscript s represents source images;
The set that all pixels are formed in i dominant color zone of expression source images,
The geometric center in i dominant color zone of expression source images,
The sum of all pixels in i dominant color zone of expression source images, I
s(x, y) expression be positioned in the source images position (x, the pixel color of y) locating,
The color average in i dominant color zone of expression source images,
The color variance in i dominant color zone of expression source images,
I dominant color zone of expression source images, M
sThe dominant color region quantity of expression source images.
The feature of calculating target image dominant color zone comprises as follows:
Subscript t represents target image in the formula;
The set that all pixels are formed in j dominant color zone of expression target image,
The geometric center in j dominant color zone of expression target image,
The sum of all pixels in j dominant color zone of expression target image, I
t(x, y) expression be positioned in the target image position (x, the pixel color of y) locating,
The color average in j dominant color zone of expression target image,
The color variance in j dominant color zone of expression target image,
J dominant color zone of expression target image; M
tThe dominant color region quantity of expression target image.
Dominant color zone mapping synoptic diagram as shown in Figure 3, in order between source images dominant color zone and target image dominant color zone, to set up mapping, the present invention utilizes earthquake distance (EMD) method (RUBNER, Y., TOMASI, C, AND GUIBAS, L.J.2000.The earth mover ' s distance as a metric for image retrieval.Int.J.Comput.Vision 40,2,99-121.) the multi-to-multi probability mapping of the best of structure between two dominant color regional ensembles.Asking for source images dominant color zone comprises the step of the mapping cost summation minimum between the dominant color zone to probability mapping optimum between the target image dominant color zone:
If the probability that i dominant color zone of source images is mapped to j dominant color zone of target image is w
Ij, the mapping cost between the dominant color zone is l
Ij, then Dui Ying cost summation is:
M
sAnd M
tBe respectively the quantity in source images and target image dominant color zone.In this formula, mapping probability w
IjNeed satisfy following constraint:
M
sBe the dominant color region quantity of source images, M
tDominant color region quantity for target image.
(Earth Mover ' s Distance) finds the solution optimum probability by the earthquake distance algorithm
Make mapping cost summation cost minimum:
For the space distribution of target image color also can be delivered in the source images, method of the present invention is at structure mapping cost function l
IjIn time, be integrated into the color distortion between the dominant color zone and differences in spatial location in the mapping cost simultaneously.This mapping cost function is defined as follows:
Wherein
Geometric center and the color average in i dominant color zone of expression source images,
Geometric center and the color average in j dominant color zone of expression target image, || .|| is for asking for Euclidean distance, ε
gAnd ε
cBe respectively the contribution factor of control geological information and colouring information, be used for controlling geometric position information and colouring information to shining upon the contribution of cost.In the method for the present invention, ε
gUsually at [1.5,2.0] interval value, ε
cUsually in [1.0,1.5] interval value.
Try to achieve above-mentioned optimum mapping probability
After, utilize this optimum mapping to recomputate color average and the color variance in dominant color zone in the source images.To each source images dominant color zone
Be defined as follows mapping:
Be the color variance in i dominant color zone of source images,
Be the color variance in j dominant color zone of target image,
I new color average and color variance in dominant color zone of expression source images.
Utilize new color average and the color variance in source images dominant color zone, can further upgrade the color of all pixels in this zone, to i dominant color zone of source images
All interior pixels are shone upon::
I
o(x is y) for being positioned at source images (x, the color after the pixel of y) the locating mapping.Yet, directly utilize the feature in single dominant color zone that the color of pixel is upgraded after, can produce tangible partitioning boundary between the dominant color zone.Tangible partitioning boundary occurs when preventing from upgrading the pixel color of source images, method of the present invention is at the color I that upgrades a certain pixel of source images
s(x in the time of y), not only utilizes the dominant color zone at this pixel place
Mapping, also combination simultaneously
The mapping in all of its neighbor dominant color zone.These are used to upgrade pixel color I
s(x, zone y) are defined as pixel I
s(x, y) in abutting connection with dominant color regional ensemble NB (x, y):
For adjacency dominant color regional ensemble NB (x, y) each the dominant color zone in
Pixel I
s(x is y) with probability
jP
XyBelong to this dominant color zone
Method of the present invention is by following formal definition probability
jP
Xy:
Z is normalized factor,
Be pixel I
s(x is y) with the dominant color zone
Similarity.In order to allow the space distribution of target image dominant color pass to source images, method of the present invention is calculated similarity measurement in conjunction with geometric similarity and color similarity
According to above-mentioned mapping of trying to achieve and probability, to each color of pixel I in the source images
s(x y) upgrades, and obtains exporting color I
o(x, y).The renewal process of color is as follows:
jP
XyBe pixel I
s(x y) belongs to j dominant color zone of source images
Probability,
With
Be j new color average and color variance in dominant color zone of source images.
The color transmittance process that method of the present invention relates to, except image was cut apart use LUV color space, all the other steps were all used CIEL
*a
*b
*Color space.When the output color of calculating pixel, can only change the tone value a of its color
*And b
*, the original brightness value L of retaining color
*Thereby, make the gradient of source images and detail content preserve.
Realize method described in the invention with C Plus Plus and opencv2.1, and used the sample image of various different colours styles to test.All experiments all are at one
Core
TM2Quad CPU Q6600 finishes on the PC of 2.40GHz 2G internal memory.
Accompanying drawing 4 has provided the source images dominant color quantity synthetic result under the target image situation.(a) among Fig. 4 is source images, and (b) among Fig. 4 is target image, the result that (c) among Fig. 4 obtains for the inventive method, the result that (d) among Fig. 4 obtains for the method for Dong.This example utilizes target image (b) to change the color of purple flowering shrubs in the source images.As can be seen from the figure, method of the present invention can all be delivered to all dominant color of target image in the source images, and dominant color has also kept its original space distribution simultaneously.
Accompanying drawing 5 has provided color transmission that source images obtains when close with target image semantic content figure as a result.(a) among Fig. 5 and (b) among Fig. 5 are respectively source images and target image, the result that (c) among Fig. 5 obtains for method of the present invention, the result that (d) among Fig. 5 obtains for the method for Dong.As can be seen, all dominant color that the image that method of the present invention produces has not only been preserved target image have also kept the fade effect of color of object simultaneously, and the part semantic information of target image also has been passed in the source images from figure as a result.
Characteristic and the innovation of method of the present invention are that the space distribution that it is corresponding also is delivered in the source images when transmitting the target image color, thereby all dominant color that guarantee target image can be kept in the result images.And by control geological information contribution factor, to the constraint strength of locus, can access the result images of different-effect when changing mapping.
Above-mentioned experimental result and the method for utilizing color space to distribute color is transmitted can be used for computer graphics and each application of Digital Image Processing, have higher actual application value.
The above; only be the embodiment among the present invention, but protection scope of the present invention is not limited thereto, anyly is familiar with the people of this technology in the disclosed technical scope of the present invention; conversion or the replacement expected can be understood, all of the present invention comprising within the scope should be encompassed in.
Claims (9)
1. method of utilizing color space to distribute color is transmitted is characterized in that comprising that step is as follows:
Step S1: combining image color of pixel and spatial positional information, utilize the average drifting algorithm that source images and target image are cut apart, the average drifting algorithm is combined into the vector of one five dimension coloured image or three-dimensional gray level image with color of pixel and spatial positional information, and utilizes the non-parametric estmation method that the probability density of vector is estimated; Each extreme point of probability density all is considered as a cluster centre; Image can be divided into several zones according to these cluster centres, be referred to as the dominant color zone;
Step S2: regard the pixel in the dominant color zone as a sample set, each dominant color zone in source images, the target image is calculated, obtain the following feature in each dominant color zone: all pixels that color average, color variance, geometric center and this dominant color zone comprise;
Step S3: color average and the geometric center information of utilizing the dominant color zone, between source images dominant color zone and target image dominant color zone, set up the probability mapping of a multi-to-multi, each dominant color zone of source images is mapped to each dominant color zone of target image with different probability, asks for source images dominant color zone and shine upon the mapping cost summation minimum that makes between the dominant color zone to probability optimum between the target image dominant color zone;
Step S4: according to the probability mapping of optimum, again source images dominant color zone is calculated, obtain new color average and the color variance in source images dominant color zone;
Step S5: utilize new color average and color variance, each color of pixel in the source images dominant color zone is upgraded.
2. by the described method of utilizing color space to distribute color is transmitted of claim 1, it is characterized in that: described dominant color zone is the contiguous pixels zone that has similar color attribute in source images or the target image.
3. it is characterized in that by the described method of utilizing the color space distribution that color is transmitted of claim 1: the feature of calculating source images dominant color zone comprises as follows:
In the formula, subscript s represents source images;
The set that all pixels are formed in i dominant color zone of expression source images,
The geometric center in i dominant color zone of expression source images,
The sum of all pixels in i dominant color zone of expression source images, I
s(x, y) expression be positioned in the source images position (x, the pixel color of y) locating,
The color average in i dominant color zone of expression source images,
The color variance in i dominant color zone of expression source images,
I dominant color zone of expression source images, M
sThe dominant color region quantity of expression source images; || .|| is for asking for Euclidean distance.
4. it is characterized in that by the described method of utilizing the color space distribution that color is transmitted of claim 3: the feature of calculating target image dominant color zone comprises as follows:
Subscript t represents target image in the formula;
The set that all pixels are formed in j dominant color zone of expression target image,
The geometric center in j dominant color zone of expression target image,
The sum of all pixels in j dominant color zone of expression target image, I
t(x, y) expression be positioned in the target image position (x, the pixel color of y) locating,
The color average in j dominant color zone of expression target image,
The color variance in j dominant color zone of expression target image,
J dominant color zone of expression target image; M
tThe dominant color region quantity of expression target image; || .|| is for asking for Euclidean distance.
5. by the described method of utilizing color space to distribute color is transmitted of claim 4, it is characterized in that: ask for source images dominant color zone and to probability mapping optimum between the target image dominant color zone step of the mapping cost summation minimum between the dominant color zone is comprised:
If the probability that i dominant color zone of source images is mapped to j dominant color zone of target image is w
Ij, the mapping cost between the dominant color zone is l
Ij, then Dui Ying cost summation cost is:
Utilize the earthquake distance algorithm to calculate optimum probability
, make mapping cost summation cost minimum be:
6. by the described method of utilizing color space to distribute color is transmitted of claim 5, it is characterized in that: calculate the mapping cost function l between the dominant color zone
IjComprise:
Wherein
Geometric center and the color average in i dominant color zone of expression source images,
Geometric center and the color average in j dominant color zone of expression target image, || .|| is for asking for Euclidean distance, ε
g, ε
cBe respectively the contribution factor of control geological information and colouring information.
7. by the described method of utilizing color space to distribute color is transmitted of claim 5, it is characterized in that: new color average and the color variance in source images dominant color zone is expressed as follows:
8. by the described method of utilizing color space to distribute color is transmitted of claim 7, it is characterized in that: to i dominant color zone of source images
All interior pixels are shone upon:
I
O(x is y) for being positioned at source images (x, the color after the pixel of y) the locating mapping.
9. by the described method of utilizing color space to distribute color is transmitted of claim 3, it is characterized in that: for the pixel in i dominant color zone of source images, except the mapping that participates in i dominant color zone, also participate in it in abutting connection with the mapping process in dominant color zone, all mapping result are done weighted mean, obtain the final color of described pixel.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110396172 CN102521807B (en) | 2011-12-02 | 2011-12-02 | Method for transferring colors by utilizing color space distribution |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110396172 CN102521807B (en) | 2011-12-02 | 2011-12-02 | Method for transferring colors by utilizing color space distribution |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102521807A CN102521807A (en) | 2012-06-27 |
CN102521807B true CN102521807B (en) | 2013-09-25 |
Family
ID=46292712
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110396172 Active CN102521807B (en) | 2011-12-02 | 2011-12-02 | Method for transferring colors by utilizing color space distribution |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102521807B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3057304A1 (en) * | 2015-02-11 | 2016-08-17 | Xiaomi Inc. | Method and apparatus for generating image filter |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102903128B (en) * | 2012-09-07 | 2016-12-21 | 北京航空航天大学 | The video image content editor's transmission method kept based on Similarity of Local Characteristic Structure |
CN103413331B (en) * | 2013-07-29 | 2016-03-30 | 北京航空航天大学 | A kind of support edits the high resolution video image content sparse expression method propagated |
CN105069465B (en) * | 2015-07-21 | 2018-06-08 | 华南农业大学 | A kind of color changeover method kept based on L0 gradients |
GB201608259D0 (en) | 2016-05-11 | 2016-06-22 | Magic Pony Technology Ltd | Feature transfer |
TW201812353A (en) * | 2016-09-13 | 2018-04-01 | 美商3M新設資產公司 | Single packet reflective polarizer with thickness profile tailored for low color at oblique angles |
US10552977B1 (en) | 2017-04-18 | 2020-02-04 | Twitter, Inc. | Fast face-morphing using neural networks |
CN107563976B (en) * | 2017-08-24 | 2020-03-27 | Oppo广东移动通信有限公司 | Beauty parameter obtaining method and device, readable storage medium and computer equipment |
CN113021898B (en) * | 2021-02-24 | 2023-03-28 | 深圳市裕同包装科技股份有限公司 | Optimization method, optimization device and optimization system for color three-dimensional printing |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI380677B (en) * | 2007-12-31 | 2012-12-21 | Altek Corp | Method of color space transformation to transfer prospect for camera sphere |
CN101441763B (en) * | 2008-11-11 | 2012-05-16 | 浙江大学 | Multiple-colour tone image unity regulating method based on color transfer |
CN101706965A (en) * | 2009-11-03 | 2010-05-12 | 上海大学 | Method for colorizing regional image on basis of Gaussian mixture model |
-
2011
- 2011-12-02 CN CN 201110396172 patent/CN102521807B/en active Active
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3057304A1 (en) * | 2015-02-11 | 2016-08-17 | Xiaomi Inc. | Method and apparatus for generating image filter |
Also Published As
Publication number | Publication date |
---|---|
CN102521807A (en) | 2012-06-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102521807B (en) | Method for transferring colors by utilizing color space distribution | |
Liu et al. | Effective template update mechanism in visual tracking with background clutter | |
US11816907B2 (en) | Systems and methods for extracting information about objects from scene information | |
Xiong et al. | Panicle-SEG: a robust image segmentation method for rice panicles in the field based on deep learning and superpixel optimization | |
Shan | Image segmentation method based on K-mean algorithm | |
CN102308306B (en) | A constraint generator for use in image segregation | |
CN104537676B (en) | Gradual image segmentation method based on online learning | |
Lavoué et al. | A new CAD mesh segmentation method, based on curvature tensor analysis | |
CN102903128B (en) | The video image content editor's transmission method kept based on Similarity of Local Characteristic Structure | |
EP1791082B1 (en) | Feature extraction using pixel cues and object cues | |
CN103632361B (en) | An image segmentation method and a system | |
US8666170B2 (en) | Computer system and method of matching for images and graphs | |
CN102982544B (en) | Many foreground object image interactive segmentation method | |
CN101802842A (en) | System and method for identifying complex tokens in an image | |
CN103617596A (en) | Image color style transformation method based on flow pattern transition | |
CN110110755B (en) | Pedestrian re-identification detection method and device based on PTGAN region difference and multiple branches | |
Liao et al. | Hdp-net: Haze density prediction network for nighttime dehazing | |
CN105590327A (en) | Motion estimation method and apparatus | |
Zhu et al. | Atmospheric light estimation in hazy images based on color-plane model | |
US20220180016A1 (en) | Determining 3D Structure Features From DSM Data | |
CN102074003A (en) | Mean shift-based embedded type image tracking system | |
Qu et al. | Parameter optimization for point clouds denoising based on no-reference quality assessment | |
Wang et al. | Multidimensional particle swarm optimization-based unsupervised planar segmentation algorithm of unorganized point clouds | |
CN106340024A (en) | Image segmentation method and application and computing device | |
Han et al. | 3D shape creation by style transfer |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |