CN101727656A - Image segmenting method based on data field - Google Patents
Image segmenting method based on data field Download PDFInfo
- Publication number
- CN101727656A CN101727656A CN200810172235A CN200810172235A CN101727656A CN 101727656 A CN101727656 A CN 101727656A CN 200810172235 A CN200810172235 A CN 200810172235A CN 200810172235 A CN200810172235 A CN 200810172235A CN 101727656 A CN101727656 A CN 101727656A
- Authority
- CN
- China
- Prior art keywords
- value
- image
- gesture
- data field
- data fields
- 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.)
- Granted
Links
Images
Abstract
The invention discloses an image segmenting method based on a data field, comprising the following steps of: defining each pixel of an image as a data object in a two-dimensional space; defining the gray value of the pixel as the quality of the data object; obtaining the potential value of each pixel by utilizing the gray value of each pixel; generating an equipotential line according to the potential value of the pixel so that an image data field is generated; and segmenting the image by utilizing the image data field. Experiments show that compared with the prior art, the image segmenting method can obtain better image segmenting effect and also has very good noise immunity.
Description
Technical field
The present invention relates to image segmentation, Target Recognition in the Flame Image Process, particularly a kind of method of utilizing data fields to realize image segmentation.
Background technology
Image segmentation just is meant the zone that image is divided into each tool characteristic, and utilizes Partial Feature in the image information to extract the technology and the process (Zhang Yujin, 2001) of some interesting targets in the image.Image segmentation is a gordian technique in the Flame Image Process, also is the subject matter in the computer vision field Level Visual, is again a classic problem simultaneously.
Data fields is the physics method (Li De is firm, Du Yi, 2005) that a kind of uncertain knowledge is found.The notion of its reference physics midfield is described the interaction between object.With the notion in the natural language, language value, speech, even the data point in the domain space, each pixel in the image, all regard interactional object or object in the space as, caing be compared to is point charge or particle in the physical field space, the every other object that is positioned at the field all is subjected to the effect of this object, form a field in whole domain space, whole number field space forms a data fields.The data fields theory is a theoretical foundation with cognitive physics, for data analysis provides a kind of brand-new thinking with handling.It is that data analysis is carried out on the basis visually with the equipotential line of physical field, and the notion of entropy is that data analysis is provided fundamental basis in the combining information opinion.
Data fields has obtained some application in Flame Image Process.As: adopting the notion of data fields and " gesture ", is research object with the human face expression image, by extracting the local maximum point of data fields different levels, rising to of real concept granularity, realize based on data fields Image mining (Dai Xiaojun, the Gan literary composition is gorgeous, Li Deyi, 2004).The pixel of facial image as data point, with the quality of grey scale pixel value as data point, is formed data fields, proposed feature extraction algorithm, and be applied to recognition of face (Lv Huijun, 2002) based on data fields.Data fields is introduced the image segmentation field realize that Study of Image Segmentation except the research work that applicant and student thereof of the present invention was engaged in, does not also have the report of other researchists' correlative study.
The process of image segmentation is actual on being is the process of all gray-scale values of each pixel of image or image being carried out cluster, and clustering result just is divided into view data the zones of different of each tool characteristic naturally.The natural topology cluster characteristic of the equipotential line of data fields provides feasibility for image segmentation, has certain advantage for utilizing data fields to carry out image segmentation.The present invention is a theoretical foundation with cognitive physics, introduces data fields and is image segmentation and effect assessment and proposed a kind of brand-new thinking and method effectively.
Experiment shows that with respect to prior art, method of the present invention can obtain the better pictures segmentation effect, and has good noise proofness.
The present invention is described in detail below in conjunction with accompanying drawing.
Summary of the invention
The purpose of this invention is to provide a kind of image partition method, utilize the natural topology cluster characteristic of the equipotential line of data fields to realize image segmentation based on data fields.
Image partition method based on data fields of the present invention may further comprise the steps:
Is each pixel definition in the image a data object in the two-dimensional space;
The gray-scale value of pixel is defined as the quality of described data object;
Utilize the gray-scale value of each pixel, obtain the gesture value of each pixel;
Equipotentiality value according to described pixel generates equipotential line, thereby generates the view data field; And
Utilize described data fields split image.
Wherein, the pixel gesture value of arbitrfary point x obtains by following potential function formula in the described data fields:
Wherein, ρ
Ij=A (i, j) (i=1,2,3 ..., m; J=1,2,3 ..., n) being grey scale pixel value, A is a gray level image, σ is a factor of influence.
Wherein, the described Interactive Segmentation of utilizing the data fields split image to be based on data fields may further comprise the steps:
By amplification, moving image data field, man-machine interactive ground selects suitable gesture value as segmentation threshold;
Utilize described segmentation threshold, the view data field is cut apart.
Wherein, described man-machine interactive ground selects suitable gesture value to comprise as the step of segmentation threshold:
According to noise or cut apart the targets of interest situation, adjust factor of influence σ, obtain the better images data fields;
By amplification, moving image data field, find the pairing gesture value of targets of interest interval (scope);
From described gesture value interval, select a suitable gesture value.
Wherein, the described data fields split image that utilizes is based on cutting apart automatically of equipotentiality value, may further comprise the steps:
From described view data field, find gesture value maximal value and minimum value;
Cut apart the class number according to what will cut apart, average from the interval of described gesture value maximal value and minimum value and cut apart value, thereby obtain automatic segmentation threshold at the view data field; And
Utilize described automatic segmentation threshold, image segmentation is cut apart automatically.
The wherein said class number of cutting apart obtains like this: according to noise with cut apart the targets of interest situation, adjust factor of influence σ, obtain the optimized image data fields; Be worth pairing pixel cluster figure according to the different gesture in the optimized image data fields, the quantity of selected pixels cluster, with this as cutting apart the class number.
Wherein, described automatic segmentation threshold comprises a plurality of different interval segmentation thresholds that are suitable for, and its quantity is corresponding to the described class number of cutting apart.
Wherein, the described data fields split image that utilizes is based on gesture value-frequency division, may further comprise the steps:
According to the equipotentiality value of described view data field, seek maximum, minimal potential value;
Calculate the frequency of each gesture value correspondence in maximum, the minimal potential value interval, generate gesture value and frequency data point one to one thus;
Removing castration value two ends frequency is zero data point, then according to these data points, generates gesture value-frequency data field;
Eliminate the gesture heart of data fields successively, generate each cluster centre;
Automatically realize image segmentation according to cluster centre.
Wherein, the step of described generation gesture value-frequency data field comprises: in the gesture value is that abscissa axis, corresponding frequencies are on the two-dimensional coordinate of axis of ordinates, makes up gesture value and frequency data point one to one, forms gesture value-frequency data field by these data points.
Description of drawings
Fig. 1 has shown the generation of view data field, and wherein Fig. 1 (a) has shown the former figure of cameraman, and Fig. 1 (b) is the view data field of cameraman image;
Fig. 2 has shown that factor of influence is to former figure of cameraman and the influence that adds the view data field, back of making an uproar, reflected by determining suitable factor of influence, can reduce The noise well, wherein Fig. 2 (a) has shown the view data field of the former figure of cameraman of factor of influence σ=0.01 o'clock, Fig. 2 (b) has shown the cameraman view data field after making an uproar of adding of factor of influence σ=0.01 o'clock, Fig. 2 (c) has shown the view data field of the former figure of cameraman of factor of influence σ=1.0 o'clock, Fig. 2 (d) has shown the cameraman view data field after making an uproar of adding of factor of influence σ=1.0 o'clock, Fig. 2 (e) has shown the view data field of the former figure of cameraman of factor of influence σ=2.0 o'clock, Fig. 2 (f) has shown the cameraman view data field after making an uproar of adding of factor of influence σ=2.0 o'clock, Fig. 2 (g) has shown the view data field of the former figure of cameraman of factor of influence σ=4.0 o'clock, Fig. 2 (h) has shown the cameraman view data field after making an uproar of adding of factor of influence σ=4.0 o'clock, Fig. 2 (i) shown the view data field of the former figure of cameraman of factor of influence σ=10.0 o'clock, and Fig. 2 (j) has shown the cameraman view data field after making an uproar of adding of factor of influence σ=10.0 o'clock;
Fig. 3 has shown the process of cutting apart based on the interactive image of data fields, wherein Fig. 3 (a) has shown the former figure of cameraman, Fig. 3 (b) has shown the original data field that is generated, Fig. 3 (c) has shown the view data field under the best factors of the original image that superposeed, Fig. 3 (d) has shown the optimum factor of influence hypograph data fields that does not have superimposed image, Fig. 3 (e) has shown the threshold value of the equipotentiality value of interactive selective extraction specific objective (hand), thereby extract interesting target, Fig. 3 (f) has shown the result who utilizes interactive approach to realize image segmentation, has extracted interesting target (hand) effectively;
Fig. 4 has shown the image segmentation process of cutting apart automatically based on the equipotentiality value, wherein Fig. 4 (a) has shown original image, Fig. 4 (b) has shown the optimum factor of influence hypograph data fields of the original image that superposeed, Fig. 4 (c) has shown the optimum factor of influence hypograph data fields of the original image that do not superpose, and Fig. 4 (d) has shown the result that several 10 automatic realizations are cut apart according to the input class;
Fig. 5 has shown the image segmentation process based on " gesture value-frequency ", wherein Fig. 5 (a) has shown the Lena original image, Fig. 5 (b) has shown the view data field that is generated, Fig. 5 (c) has shown " gesture value-frequency " data point that is generated, Fig. 5 (d) has shown the two dimension that is generated " gesture value-frequency " data fields, Fig. 5 (e) has shown the three-dimensional that is generated " gesture value-frequency " data fields, Fig. 5 (f) has shown the result who eliminates the gesture heart for the first time, Fig. 5 (g) has shown the result who eliminates the gesture heart for the second time, Fig. 5 (h) has shown the result who eliminates the gesture heart for the third time, and Fig. 5 (i) has shown the result who realizes image segmentation according to cluster centre;
Fig. 6 has shown the former figure of experiment that compares experiment with other method, and wherein Fig. 6 (a) has shown the Polygon image, and Fig. 6 (b) has shown the Flower image, and Fig. 6 (c) has shown that adding the back signal to noise ratio (S/N ratio) of making an uproar is 26 Fingerprint image.
Embodiment
Image partition method based on data fields of the present invention comprises the generation of view data field, based on the Interactive Segmentation of data fields, based on the cutting apart automatically of equipotentiality value, based on gordian techniquies such as cutting apart of " gesture value-frequency ".(1) generation of view data field
The gray level image A of given m * n pixel
M * n, as if a data object that each pixel is considered as in the two-dimensional space, with the gray scale ρ of pixel
Ij=A (i, j) (i=1,2,3 ..., m; J=1,2,3 ..., n) quality that is considered as data object (is supposed gray-scale value ρ
IjNormalizing is to [0,1] interval), then the interaction of all pixels in the 2-D data space just can be determined a view data field.According to the potential function formula of data fields, the gesture value of any point x may be calculated in
Calculate the gesture value of each image pixel according to above formula, generate equipotential line according to the gesture value of image pixel, thereby generate the view data field.
(2) based on the Interactive Segmentation of data fields
Behind the generation method generation view data field according to the view data field, by amplification, moving image data field, man-machine interaction ground selects suitable gesture value as segmentation threshold, thereby extracts interested target.Specifically may further comprise the steps:, generate the data fields of this image according to the method for being introduced in the first step (generation of view data field) 1) for a width of cloth gray level image; 2) according to noise or cut apart target conditions, adjust factor of influence, regenerate the view data field; 3) by amplification, moving image data field, the pairing gesture value of object observing interval; 4) segmentation threshold is set, the view data field is cut apart; 5) initial results of acquisition image segmentation; 6) observe segmentation result,, continue to adjust factor of influence, cut apart again, till meeting the demands if undesirable.
(3) cutting apart automatically based on the equipotentiality value
Gordian technique (2) but in the advantage of the interactive image dividing method introduced be that parameter is adjusted to obtain satisfied segmentation result in man-machine interaction ground in cutting apart.Yet, too many parameter input and the subjectivity of definite threshold value be a drawback of this method.Therefore, on the basis of view data field, can cut apart according to equipotentiality value realization automated graphics.Its basic ideas are: find the maximal value and the minimum value of view data field, average and cut apart value according to cutting apart the class number, thereby obtain the automatic segmentation threshold at the view data field, thereby realize image segmentation.Specifically may further comprise the steps: 1) generate the view data field; 2) according to noise situations with cut apart purpose, adjust factor of influence, obtain the optimized image data fields; 3) determine to cut apart the class number, input correlation parameter (cutting apart the class number); 4) realize image segmentation according to the class number of being imported of cutting apart.
(4) cutting apart based on " gesture value-frequency "
According to the view data field, structure has the data point of two dimension attributes, is that horizontal ordinate, corresponding frequencies are ordinate with the gesture value.By these dot generation " gesture value-frequency " data fields, make its natural cluster, both considered the distance of gesture codomain, considered the gap of frequency field again.By eliminating the method for the gesture heart, find cluster centre successively.According to the gesture value of cluster centre correspondence, carry out automated graphics and cut apart.The thought that Here it is cuts apart based on the automated graphics of " gesture value-frequency data field ".This method specifically may further comprise the steps: 1) generate the view data field; 2) according to the equipotentiality value of view data field, seek maximum, minimal potential value; 3) according to minimax gesture value, calculate corresponding frequency, generate " gesture value-frequency " data point; 4) data point is put in order, removing castration value two ends frequency is zero data point, then according to these data points, generates " gesture value-frequency " data fields; 5) eliminate the gesture heart of data fields successively, generate each cluster centre; 6) realize image segmentation automatically according to cluster centre.
Below by an embodiment, describe the image partition method based on data fields of the present invention in detail.
Image partition method based on data fields of the present invention may further comprise the steps:
Is each pixel definition in the image a data object in the two-dimensional space;
The gray-scale value of pixel is defined as the quality of described data object;
Utilize the gray-scale value of each pixel, obtain the gesture value of each pixel;
Equipotentiality value according to described pixel generates equipotential line, thereby generates the view data field; And
Utilize described data fields split image.
Here can according to the gesture value from big to small or trend from small to large generate equipotential line.
The pixel gesture value of arbitrfary point x obtains by following potential function formula in the wherein said data fields:
Wherein, ρ
Ij=A (i, j) (i=1,2,3 ..., m; J=1,2,3 ..., n) being grey scale pixel value, A is a gray level image, σ is a factor of influence.
Utilize the data fields split image to be based on the Interactive Segmentation of data fields, may further comprise the steps:
By amplification, moving image data field, man-machine interactive ground selects suitable gesture value as segmentation threshold;
Utilize described segmentation threshold, the view data field is cut apart.
The view data field is according to the interaction of image slices vegetarian refreshments in the 2-D data space, utilizes data fields potential function computing formula
Calculate the gesture value of each pixel position, the gesture value has reflected the interactional size of this pixel and surrounding pixel.Image segmentation is to segment the image into the zone that each has similar characteristic, has similar characteristic between the pixel of every kind of image-region inside, has interaction between pixel, therefore can utilize the gesture value of view data field to reflect this interactional size.Therefore, can realize image segmentation according to this interactional size.
Because the view data field comprises the interval with different gesture values, therefore utilize segmentation threshold, the view data field is cut apart, be exactly to extract and the corresponding targets of interest of this segmentation threshold scope.With Fig. 1 is example, if purpose is in order to extract " hand ", and by amplification local data field, the gesture value interval of object observing " hand ", by observing, setting threshold is 52.068142, thereby extracts the target " hand " of these threshold value 52.068142 ± 3.06 scopes.
Wherein, described man-machine interactive ground selects suitable gesture value to comprise as the step of segmentation threshold: according to noise or cut apart the targets of interest situation, adjust factor of influence σ, obtain the better images data fields; By amplification, moving image data field, find the pairing gesture value of targets of interest interval; From described gesture value interval, select a suitable gesture value.
Wherein, the described data fields split image that utilizes is based on cutting apart automatically of equipotentiality value, may further comprise the steps:
From described view data field, find gesture value maximal value and minimum value;
Cut apart the class number according to what will cut apart, average from the interval of described gesture value maximal value and minimum value and cut apart value, thereby obtain automatic segmentation threshold at the view data field; And
Utilize described automatic segmentation threshold, image is cut apart automatically.
Realize that according to the view data field the main thought of image segmentation is exactly to determine different image classification zones according to view data field potential value size, therefore must find out the maximal value and the minimum value of view data field potential value, interval according to maximal value and minimum specified data field potential value, then cluster being carried out in this interval divides, be divided into each subclass, corresponding image-region of each subclass and image type, thus realize image segmentation.Described here utilize described automatic segmentation threshold, image is cut apart automatically, extract its gesture value and segmentation threshold or the corresponding subclass of its scope automatically exactly.
For image segmentation, the gradation of image interval that each image-region and image type are corresponding certain, thereby the gesture value interval of corresponding certain view data field.In the image segmentation process, specifically being divided into what zones can be determined according to actual needs by the user.For example: for certain image, the user need extract the image object that specifies number, and according to the number of the image object of appointment data field potential value scope is averaged division, can realize that automated graphics cuts apart.
Wherein, the described class number of cutting apart obtains by following steps:
According to noise with cut apart the targets of interest situation, adjust factor of influence σ, obtain the optimized image data fields;
Be worth pairing pixel cluster figure according to the different gesture in the optimized image data fields, the quantity of selected pixels cluster, with this as cutting apart the class number.
Wherein, described automatic segmentation threshold comprises a plurality of different interval segmentation thresholds that are suitable for, and its quantity is corresponding to the described class number of cutting apart.
Wherein, the described data fields split image that utilizes is based on gesture value-frequency division, may further comprise the steps:
According to the equipotentiality value of described view data field, seek maximum, minimal potential value;
Calculate the frequency of each gesture value correspondence in maximum, the minimal potential value interval, generate gesture value and frequency data point one to one thus;
Removing castration value two ends frequency is zero data point, then according to these data points, generates gesture value-frequency data field;
Eliminate the gesture heart of data fields successively, generate each cluster centre;
Automatically realize image segmentation according to cluster centre.
Determined that after the gesture value maximal value and minimum value interval of view data field, each gesture of adding up this interval respectively is worth the number of pairing pixel,, just can obtain having the frequency of the pixel of this gesture value then divided by the total number of pixel.
For example: for gesture value maximal value in and minimum value between certain gesture value, as: 52.0, add up the pixel number that has this gesture value in the whole image data field respectively, remove then in total pixel number, obtain having the frequency of this gesture value.
As can be seen, the extreme value place of potential field has often been represented the cluster centre of some data, because these local extremums have embodied the overall permanence of local data from data fields.Find the extreme value of these gesture values in the data fields, also just be equal to the cluster centre that has found these data.The gesture heart of so-called data fields just is meant the cluster centre of these data fields.
The method of eliminating the gesture heart of data fields can be: before obtaining next cluster centre point, be necessary to eliminate the influence of debating the cluster centre point of knowing just now.Be to debate the cluster centre point known to the influence of cluster once more in order to reject, so the focus of revising is the data radiation radius, making the gesture value of debating the cluster gesture heart point of knowing is 0.After eliminating the gesture heart, all should recomputate the field intensity of data fields once more at every turn.Therefore, the new field majorant that obtains after the modification is a recurrence function formula.
Generating each cluster centre can realize like this: at first find out first cluster centre, utilize be eliminated data fields behind the gesture heart of the method for the aforesaid elimination data field potential heart then, look for cluster centre once more from the data fields of eliminating behind the gesture heart then, obtain second cluster centre; And then the data fields of method that utilize to eliminate the gesture heart after being eliminated, and then look for cluster centre, obtain the 3rd cluster centre like this.The rest may be inferred obtains each cluster centre.
By cluster is carried out in the view data field, obtain different clusters, each cluster has a cluster centre, determine different clusters according to cluster centre, corresponding image type of each class and image-region, give a kind of image type to the image pixel that belongs to each cluster, thereby realize image segmentation.
Wherein, the step of described generation gesture value-frequency data field comprises: in the gesture value is that abscissa axis, corresponding frequencies are on the two-dimensional coordinate of axis of ordinates, makes up gesture value and frequency data point one to one, forms gesture value-frequency data field by these data points.
Below by several examples, further specify.
Example 1: Fig. 1 has shown the generation of view data field.
Example 2: Fig. 2 has shown that factor of influence to former figure of cameraman and the influence that adds the view data field, back of making an uproar, has reflected by determining suitable factor of influence, can reduce The noise well.
Example 3: Fig. 3 has shown the process of cutting apart based on the interactive image of data fields.Experimental result shows that this method has obtained effect preferably, can extract interesting target (hand) well.
Example 4: Fig. 4 has shown the image segmentation process of cutting apart automatically based on the equipotentiality value.Experiment shows that this method can realize cutting apart automatically of image quickly and easily.
Example 5: Fig. 5 has shown the image segmentation process based on " gesture value-frequency ".Experiment shows that this method has obtained good segmentation effect.
Example 6: choose three width of cloth experimental image as shown in Figure 6: Polygon image, Flower image and the adding back signal to noise ratio (S/N ratio) of making an uproar is that 26 Fingerprint image compares experiment, will be based on the image partition method of the equipotentiality value of data fields and the comparative analysis that optimal threshold split plot design, fuzzy C-means clustering FCM image method are missed the branch rate, the comparative analysis result is as shown in table 1.
Table 1: the mistake branch rate contrast of different images dividing method
As can be seen from Table 1, generally, these three kinds of methods all compare effectively.To the simplest Polygon image, these three kinds of algorithms all show good performance, but poor slightly based on the equipotentiality value dividing method of data fields; To comparatively complicated Flower image, more approaching on these three kinds of method performances; And, show to have shown its advantage again based on the method for data fields to the Fingerpinter image of noise spot is arranged.Show that the data fields image partition method has good noise proofness.
Although above the present invention is had been described in detail, the invention is not restricted to this, those skilled in the art of the present technique can carry out various modifications according to principle of the present invention.Therefore, all modifications of doing according to the principle of the invention all should be understood to fall into protection scope of the present invention.
Claims (10)
1. image partition method based on data fields comprises:
At first generate the view data field;
Carry out image Interactive Segmentation then, perhaps carry out Image Automatic Segmentation, perhaps carry out image segmentation based on " gesture value-frequency " based on the equipotentiality value based on data fields.
2. image partition method based on data fields may further comprise the steps:
Is each pixel definition in the image a data object in the two-dimensional space;
The gray-scale value of pixel is defined as the quality of described data object;
Utilize the gray-scale value of each pixel, obtain the gesture value of each pixel;
Equipotentiality value according to described pixel generates equipotential line, thereby generates the view data field; And
Utilize described data fields split image.
3. image partition method according to claim 1 and 2, the pixel gesture value of arbitrfary point x obtains by following potential function formula in the wherein said data fields:
Wherein, ρ
Ij=A (i, j) (i=1,2,3 ..., m; J=1,2,3 ..., n) being grey scale pixel value, A is a gray level image, σ is a factor of influence.
4. according to claim 1 or 2 or 3 described image partition methods, the wherein said Interactive Segmentation of utilizing the data fields split image to be based on data fields may further comprise the steps:
By amplification, moving image data field, man-machine interactive ground selects suitable gesture value as segmentation threshold;
Utilize described segmentation threshold, the view data field is cut apart.
5. image partition method according to claim 4, wherein said man-machine interactive ground select suitable gesture value to comprise as the step of segmentation threshold:
According to noise or cut apart the targets of interest situation, adjust factor of influence σ, obtain the better images data fields;
By amplification, moving image data field, find the pairing gesture value of targets of interest interval (scope);
From described gesture value interval, select a suitable gesture value.
6. according to claim 1 or 2 or 3 described image partition methods, the wherein said data fields split image that utilizes is based on cutting apart automatically of equipotentiality value, may further comprise the steps:
From described view data field, find gesture value maximal value and minimum value;
Cut apart the class number according to what will cut apart, average from the interval of described gesture value maximal value and minimum value and cut apart value, thereby obtain automatic segmentation threshold at the view data field; And
Utilize described automatic segmentation threshold, image segmentation is cut apart automatically.
7. image partition method according to claim 6, the wherein said class number of cutting apart obtains by following steps:
According to noise with cut apart the targets of interest situation, adjust factor of influence σ, obtain the optimized image data fields;
Be worth pairing pixel cluster figure according to the different gesture in the optimized image data fields, the quantity of selected pixels cluster, with this as cutting apart the class number.
8. image partition method according to claim 6, wherein said automatic segmentation threshold comprise a plurality of different interval segmentation thresholds that are suitable for, and its quantity is corresponding to the described class number of cutting apart.
9. according to claim 1 or 2 or 3 described image partition methods, the wherein said data fields split image that utilizes is based on gesture value-frequency division, may further comprise the steps:
According to the equipotentiality value of described view data field, seek maximum, minimal potential value;
Calculate the frequency of each gesture value correspondence in maximum, the minimal potential value interval, generate gesture value and frequency data point one to one thus;
Removing castration value two ends frequency is zero data point, then according to these data points, generates gesture value-frequency data field;
Eliminate the gesture heart of data fields successively, generate each cluster centre;
Automatically realize image segmentation according to cluster centre.
10. image partition method according to claim 9, the step of wherein said generation gesture value-frequency data field comprises: in the gesture value is that abscissa axis, corresponding frequencies are on the two-dimensional coordinate of axis of ordinates, make up gesture value and frequency data point one to one, form gesture value-frequency data field by these data points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101722353A CN101727656B (en) | 2008-10-31 | 2008-10-31 | Image segmenting method based on data field |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101722353A CN101727656B (en) | 2008-10-31 | 2008-10-31 | Image segmenting method based on data field |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101727656A true CN101727656A (en) | 2010-06-09 |
| CN101727656B CN101727656B (en) | 2012-03-28 |
Family
ID=42448514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008101722353A Expired - Fee Related CN101727656B (en) | 2008-10-31 | 2008-10-31 | Image segmenting method based on data field |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101727656B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104067308A (en) * | 2012-01-25 | 2014-09-24 | 英特尔公司 | Object selection in an image |
| CN105023006A (en) * | 2015-08-05 | 2015-11-04 | 西安电子科技大学 | Face recognition method based on enhanced nonparametric margin maximization criteria |
| CN105976385A (en) * | 2016-05-19 | 2016-09-28 | 浙江工商大学 | Image segmentation method based on image data field |
| CN106204588A (en) * | 2016-07-08 | 2016-12-07 | 腾讯科技（深圳）有限公司 | A kind of image processing method and device |
| CN106981084A (en) * | 2016-10-28 | 2017-07-25 | 阿里巴巴集团控股有限公司 | A kind of method and device of drawing isoline |
| CN109144442A (en) * | 2018-07-13 | 2019-01-04 | 北京明牌科技有限公司 | A kind of intelligent cloud blank |
| CN113761083A (en) * | 2020-06-01 | 2021-12-07 | 中国人民解放军战略支援部队信息工程大学 | Image generation method and device based on country and geography relation |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100553431B1 (en) * | 2003-04-21 | 2006-02-20 | 주식회사 팬택 | Method for concluding threshold for image division |
| CN100576246C (en) * | 2007-05-23 | 2009-12-30 | 华中科技大学 | A kind of image partition method based on attribute histogram |
-
2008
- 2008-10-31 CN CN2008101722353A patent/CN101727656B/en not_active Expired - Fee Related
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104067308A (en) * | 2012-01-25 | 2014-09-24 | 英特尔公司 | Object selection in an image |
| CN105023006A (en) * | 2015-08-05 | 2015-11-04 | 西安电子科技大学 | Face recognition method based on enhanced nonparametric margin maximization criteria |
| CN105023006B (en) * | 2015-08-05 | 2018-05-04 | 西安电子科技大学 | Face identification method based on enhanced nonparametric maximal margin criterion |
| CN105976385A (en) * | 2016-05-19 | 2016-09-28 | 浙江工商大学 | Image segmentation method based on image data field |
| CN105976385B (en) * | 2016-05-19 | 2019-05-10 | 浙江工商大学 | A kind of image partition method based on image data field |
| CN106204588A (en) * | 2016-07-08 | 2016-12-07 | 腾讯科技（深圳）有限公司 | A kind of image processing method and device |
| CN106204588B (en) * | 2016-07-08 | 2020-03-17 | 腾讯科技（深圳）有限公司 | Image processing method and device |
| CN106981084A (en) * | 2016-10-28 | 2017-07-25 | 阿里巴巴集团控股有限公司 | A kind of method and device of drawing isoline |
| CN106981084B (en) * | 2016-10-28 | 2020-11-06 | 创新先进技术有限公司 | Method and device for drawing contour line |
| CN109144442A (en) * | 2018-07-13 | 2019-01-04 | 北京明牌科技有限公司 | A kind of intelligent cloud blank |
| CN113761083A (en) * | 2020-06-01 | 2021-12-07 | 中国人民解放军战略支援部队信息工程大学 | Image generation method and device based on country and geography relation |
| CN113761083B (en) * | 2020-06-01 | 2023-12-01 | 中国人民解放军战略支援部队信息工程大学 | Image generation method and device based on national earth relation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101727656B (en) | 2012-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101727656B (en) | Image segmenting method based on data field | |
| Panwar et al. | Image Segmentation using K-means clustering and Thresholding | |
| CN103455991A (en) | Multi-focus image fusion method | |
| CN113762138B (en) | Identification method, device, computer equipment and storage medium for fake face pictures | |
| CN104616274A (en) | Algorithm for fusing multi-focusing image based on salient region extraction | |
| CN104376556A (en) | Rock CT image target segmentation method | |
| JP5939047B2 (en) | Text enhancement method and apparatus, and text extraction method and apparatus | |
| Gritzman et al. | Comparison of colour transforms used in lip segmentation algorithms | |
| JP2020501206A (en) | Online cartoon identification system and method based on region of interest | |
| CN102081799A (en) | Method for detecting change of SAR images based on neighborhood similarity and double-window filtering | |
| CN104992176B (en) | A kind of Chinese character extracting method towards an inscription on a tablet | |
| CN106846321A (en) | A kind of image partition method based on Bayesian probability Yu neutral net | |
| JP2013164832A (en) | Object recognition device and method | |
| González-Hidalgo et al. | A new edge detector based on uninorms | |
| CN108241874A (en) | Video text area positioning method based on BP neural network and spectrum analysis | |
| CN101873506A (en) | Image processing method for providing depth information and image processing system thereof | |
| Rao et al. | Statistical analysis for performance evaluation of image segmentation quality using edge detection algorithms | |
| Singh et al. | High Density Impulse Noise Detection using Fuzzy C-means Algorithm. | |
| Noreen et al. | Hand detection using HSV model | |
| Yan et al. | Infrared image segment and fault location for power equipment | |
| CN113361321A (en) | Infrared small target detection method and device | |
| Yang et al. | An adaptive digital camouflage scheme using visual perception and k-mean clustering | |
| Xia et al. | Fast threshold selection algorithm of infrared human images based on two-dimensional fuzzy Tsallis entropy | |
| Idrus et al. | Swarm Plot: Data Redistribution in Non-Random Technique | |
| Alkishri et al. | Evaluating the Effectiveness of a Gan Fingerprint Removal Approach in Fooling Deepfake Face Detection |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120328 Termination date: 20181031 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |