CN106384369A - Data guiding color manifold obtaining method - Google Patents
Data guiding color manifold obtaining method Download PDFInfo
- Publication number
- CN106384369A CN106384369A CN201610792169.4A CN201610792169A CN106384369A CN 106384369 A CN106384369 A CN 106384369A CN 201610792169 A CN201610792169 A CN 201610792169A CN 106384369 A CN106384369 A CN 106384369A
- Authority
- CN
- China
- Prior art keywords
- color
- dimensional
- manifold
- point
- data
- 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
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a data guiding color manifold obtaining method, and the method is used to obtain a one-dimensional line or a two-dimensional plane that is most fit to image sampling color points of a specific object from a three-dimensional color space. The method comprises the following steps that 1) a two-dimensional image set of the specific object is obtained, and sampling data points of the two-dimensional images are obtained after that interference is removed; 2) the density of the sampling data points is estimated, a color histogram is established, and a characteristic color and corresponding characteristic color points are selected; and 3) the characteristic color points are fit via linear and nonlinear methods, and the one-dimensional line or two-dimensional plane is obtained. Compared with the prior art, the method of the invention has the advantages that the dimension of data is reduced in a visual and concrete way.
Description
Technical field
The present invention relates to image processing field, especially relate to a kind of color manifold acquisition methods of data guiding.
Background technology
Manifold is the space that local has Euclidean space property, and manifold is used for describing geometrical body, thing in mathematics
In reason, the four-dimensional pseudo-Riemannian manifold of the space-time model of the phase space of classical mechanics and construction general theory of relativity is all the reality of manifold
Example.
The existing low-dimensional color manifold being obtained by color manifold remains the main component color of things, in color choosing
Select, change Color Style etc. and be all widely used.
But existing color manifold is huge often with millions of color points, data during producing, and deals with
Complicated and lose time, and unnecessary characterization can be produced.
Content of the invention
The purpose of the present invention is exactly to provide a kind of Data Dimensionality Reduction, directly perceived to overcome the defect that above-mentioned prior art exists
The color manifold acquisition methods of specific data guiding.
The purpose of the present invention can be achieved through the following technical solutions：
A kind of color manifold acquisition methods of data guiding, matching can specify thing in order to obtain in three-dimensional color space
The one dimensional line of thing picture sample color point or two-dimensional surface, the method comprises the following steps：
1) obtain the two dimensional image set specifying things, after removing interference, obtain the sampled data points of two dimensional image；
2) density estimation is carried out to sampled data points, set up color histogram, choose it and characterize color and corresponding sign
Color point；
3) it is fitted to characterizing color point by linear and nonlinear method, obtain and finally take one dimensional line or two-dimensional surface.
Described step 1) in removal disturb as removing ambient interferences, concrete grammar is as follows：
After two dimensional image is carried out bilateral filtering obfuscation, judged in two dimensional image according to CIEDE2000 color distortion
Whether marginal point is connected with other color points, if connection, judges that this color point, as background, is removed.
Described step 2) specifically include following steps：
21) using RGB color and be divided into the interval of 16*16*16, setting up color histogram；
22) choose front 15% interval containing most color points as sign color point.
Described step 3) in, linear method includes PCA, and nonlinear method includes self organizing neural network
Method.
Described PCA specifically includes following steps：
Calculate color vector covariance matrix, carry out feature decomposition and obtain the as throwing of the corresponding characteristic vector of eigenvalue of maximum
Shadow vector, obtains final product one dimensional line or two-dimensional surface after being projected.
Described self organizing neural network method specifically includes following steps：
For given priming color manifold, make color manifold non-linear close to sign number of colours by neural metwork training
Strong point, finally gives one dimensional line or two-dimensional surface.
Compared with prior art, the present invention has advantages below：
First, Data Dimensionality Reduction：By color histogram, color is counted, obtain each color in picture corresponding close
Degree, chooses front 15% as characterizing color, so the color point of original million ranks is reduced to the quantity of only hundred ranks, greatly
Reduce greatly the quantity of data point, be easy to statistical computation.
2nd, intuitively concrete：By the way of manifold, the color density of picture is shown in three dimensions, Neng Gouzhi
The distribution of color situation of the embodiment picture seen.
Brief description
Fig. 1 is method of the present invention flow chart.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
Embodiment：
As shown in figure 1, the target of this method is by the image collection sample color point to specified things, and in three-dimensional face
Finding in the colour space can one-dimensional straight (bent) line of these sample color points of matching and flat (bent) face of two dimension.Thus obtained low-dimensional
Color manifold remains the main component color of things, in color selecting, changes Color Style etc. and is all widely used.Specifically
Method And Principle can be divided into three parts：
Data acquisition：The input data of this method is to search for the X-Y scheme image set that concrete things (as apple) obtains from network
Close.Image due to obtaining comprises background, and this method is used a fairly simple method and eliminated ambient interferences, by image
After carrying out bilateral filtering obfuscation, carry out judging whether two color points connect according to CIEDE2000 color distortion.Last institute
The color connecting with marginal point point is had to be considered as background.After removing background dot, sample millions of from all data points
As input data.
Density estimation：Obtain the data point through over-sampling for the previous step as input data after, RGB color is drawn
It is divided into the interval of 16*16*16, sets up color histogram, referred to as color density, take front 15% to contain most color points afterwards
Interval characterizes as color point, and corresponding quantity is the density of this color point, and so million orders of magnitude from input are compressed to
Hundreds of order of magnitude, each color point is to the significance level that a density value should be had to represent this color point.
Dimension declines：After color point in obtaining hundreds of three dimensions, the side by linear processes for this method
Formula obtains one-dimensional straight (bent) line and flat (bent) face of two dimension that can characterize native color point.Linear method employs principal component analysis
Method, by calculating face
Color vector covariance matrix, carry out feature decomposition obtain the corresponding characteristic vector of eigenvalue of maximum be projection to
Amount.The straight line so finding and the variance of plane maximization subpoint.Nonlinear method adopts self organizing neural network method, gives
Determine priming color manifold, make color manifold non-linear close to raw data points by neural metwork training, finally give one
Curve or curved surface.
Because this method is a kind of method of data guiding, so experimental result becomes dependent upon the quality of data, due to all
Data all obtains from network, the unavoidable error that there are some and lead to because of ambiguousness, so, in order to improve experimental result,
This method to the image zooming-out color characteristic in original input picture set and clusters, and retains in the cluster result obtaining
The maximum classification of accounting, removes the error that some lead to because network data is asymmetric, is that the generation of subsequent color manifold improves
The degree of accuracy.
Claims (6)
1. a kind of color manifold acquisition methods of data guiding, matching can specify things in order to obtain in three-dimensional color space
The one dimensional line of picture sample color point or two-dimensional surface are it is characterised in that the method comprises the following steps：
1) obtain the two dimensional image set specifying things, after removing interference, obtain the sampled data points of two dimensional image；
2) density estimation is carried out to sampled data points, set up color histogram, choose it and characterize color and corresponding sign color
Point；
3) it is fitted to characterizing color point by linear and nonlinear method, obtain and finally take one dimensional line or two-dimensional surface.
2. a kind of color manifold acquisition methods of data guiding according to claim 1 are it is characterised in that described step
1) removal in is disturbed as removing ambient interferences, and concrete grammar is as follows：
After two dimensional image is carried out bilateral filtering obfuscation, edge in two dimensional image is judged according to CIEDE2000 color distortion
Whether point is connected with other color points, if connection, judges this color point as background, is removed.
3. a kind of color manifold acquisition methods of data guiding according to claim 1 are it is characterised in that described step
2) following steps are specifically included：
21) using RGB color and be divided into the interval of 16*16*16, setting up color histogram；
22) choose front 15% interval containing most color points as sign color point.
4. a kind of color manifold acquisition methods of data guiding according to claim 1 are it is characterised in that described step
3) in, linear method includes PCA, and nonlinear method includes self organizing neural network method.
5. a kind of color manifold acquisition methods of data guiding according to claim 4 are it is characterised in that described main one-tenth
Point analytic approach specifically includes following steps：
Calculate color vector covariance matrix, carry out feature decomposition obtain the corresponding characteristic vector of eigenvalue of maximum be projection to
Amount, obtains final product one dimensional line or two-dimensional surface after being projected.
6. a kind of data guiding according to claim 4 color manifold acquisition methods it is characterised in that described from group
Knit neural net method and specifically include following steps：
For given priming color manifold, make color manifold non-linear close to sign color data by neural metwork training
Point, finally gives one dimensional line or two-dimensional surface.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610792169.4A CN106384369A (en) | 2016-08-31 | 2016-08-31 | Data guiding color manifold obtaining method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610792169.4A CN106384369A (en) | 2016-08-31 | 2016-08-31 | Data guiding color manifold obtaining method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106384369A true CN106384369A (en) | 2017-02-08 |
Family
ID=57938867
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610792169.4A Pending CN106384369A (en) | 2016-08-31 | 2016-08-31 | Data guiding color manifold obtaining method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106384369A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113808141A (en) * | 2020-06-15 | 2021-12-17 | 株式会社岛津制作所 | Imaging quality analysis apparatus and imaging quality analysis method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024262A (en) * | 2011-01-06 | 2011-04-20 | 西安电子科技大学 | Method for performing image segmentation by using manifold spectral clustering |
| CN102945549A (en) * | 2012-10-15 | 2013-02-27 | 山东大学 | Shot segmentation method based on manifold learning |
| CN105447884A (en) * | 2015-12-21 | 2016-03-30 | 宁波大学 | Objective image quality evaluation method based on manifold feature similarity |
-
2016
- 2016-08-31 CN CN201610792169.4A patent/CN106384369A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102024262A (en) * | 2011-01-06 | 2011-04-20 | 西安电子科技大学 | Method for performing image segmentation by using manifold spectral clustering |
| CN102945549A (en) * | 2012-10-15 | 2013-02-27 | 山东大学 | Shot segmentation method based on manifold learning |
| CN105447884A (en) * | 2015-12-21 | 2016-03-30 | 宁波大学 | Objective image quality evaluation method based on manifold feature similarity |
Non-Patent Citations (3)
| Title |
|---|
| CHUONG H. NGUYEN、TOBIAS RITSCHEL: "Data-Driven Color Manifolds", 《ACM TRANSACTIONS ON GRAPHICS》 * |
| 圣少友 等: "基于流形学习的舌图像颜色特征提取", 《航天医学与医学工程》 * |
| 沈洋 等: "交互式前景抠图技术综述", 《计算机辅助设计与图形学学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113808141A (en) * | 2020-06-15 | 2021-12-17 | 株式会社岛津制作所 | Imaging quality analysis apparatus and imaging quality analysis method |
| CN113808141B (en) * | 2020-06-15 | 2024-05-28 | 株式会社岛津制作所 | Imaging quality analysis device and imaging quality analysis method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109872285B (en) | Retinex low-illumination color image enhancement method based on variational constraint | |
| CN109522956B (en) | Low-rank discriminant feature subspace learning method | |
| CN108717524B (en) | Gesture recognition system based on double-camera mobile phone and artificial intelligence system | |
| CN110245593A (en) | A kind of images of gestures extraction method of key frame based on image similarity | |
| CN109685749A (en) | Image style conversion method, device, equipment and computer storage medium | |
| CN109829924B (en) | Image quality evaluation method based on principal feature analysis | |
| CN101833664A (en) | Video image character detecting method based on sparse expression | |
| CN110930411B (en) | Human body segmentation method and system based on depth camera | |
| CN101986295B (en) | Image clustering method based on manifold sparse coding | |
| Čuljak et al. | Classification of art paintings by genre | |
| CN106651741B (en) | graphics processing system based on cloud computing | |
| CN109117860A (en) | A kind of image classification method based on subspace projection and dictionary learning | |
| CN115641583B (en) | Point cloud detection method, system and medium based on self-supervision and active learning | |
| CN105912739B (en) | A kind of similar pictures searching system and its method | |
| CN110458792A (en) | Method and device for evaluating quality of face image | |
| CN110049242A (en) | A kind of image processing method and device | |
| CN110990617B (en) | Picture marking method, device, equipment and storage medium | |
| CN109241932B (en) | Thermal infrared human body action identification method based on motion variance map phase characteristics | |
| CN109543525B (en) | Table extraction method for general table image | |
| CN106780333A (en) | A kind of image super-resolution rebuilding method | |
| CN113963193A (en) | Method and device for generating vehicle body color classification model and storage medium | |
| CN108090914A (en) | Color image segmentation method based on statistical modeling and pixel classifications | |
| CN106384369A (en) | Data guiding color manifold obtaining method | |
| CN107818579B (en) | Color texture feature extraction method based on quaternion Gabor filtering | |
| CN113850748A (en) | Point cloud quality evaluation system and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20170208 |