CN106327507B - A kind of color image conspicuousness detection method based on background and foreground information - Google Patents
A kind of color image conspicuousness detection method based on background and foreground information Download PDFInfo
- Publication number
- CN106327507B CN106327507B CN201610654316.1A CN201610654316A CN106327507B CN 106327507 B CN106327507 B CN 106327507B CN 201610654316 A CN201610654316 A CN 201610654316A CN 106327507 B CN106327507 B CN 106327507B
- Authority
- CN
- China
- Prior art keywords
- super
- background
- pixel block
- significance
- foreground
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
Abstract
A series of color image conspicuousness detection method based on background and foreground information that the invention discloses a kind of, comprising the following steps: over-segmentation is carried out to the color image of input and handles to obtain super-pixel block；Background seed is chosen, coarse significance is obtained by Characteristic Contrast between each super-pixel block and background seed；Feature distribution based on background seed defines the background weight of each super-pixel block, improves coarse significance by background weight and obtains the significance based on background information；The notable figure based on background information formed to previous step is split, and a close foreground area is chosen in all segmentation results, extracts foreground area feature, obtains the significance based on foreground information by Characteristic Contrast；The significance based on background and foreground information of first two steps acquisition is integrated, and carries out smooth operation and obtains the significance after all super-pixel block optimizations.The present invention can protrude the foreground target in image more consistently, and have good inhibitory effect to ambient noise in image.
Description
Technical field
The invention belongs to the conspicuousness detection technique fields of image scene, and in particular to one kind is based on background and foreground information
Color image conspicuousness detection method.
Background technique
Vision significance is the important research content of visual cognition and scene understanding, is related to cognitive psychology, cognition neural
Multiple subjects such as science, computer vision.Due to the difference on the foreground target and its background existing characteristics in scene, human eye view
Feel system tends to the information of the foreground area in quick positioning scene and the priority processing region.In order to simulate human eye
This high efficiency information processing method, conspicuousness detect the extensive concern for having obtained related fields scholars in recent years.
Feature integration theory (the Anne of Treisman et al. proposition can be traced in the research of conspicuousness detection earliest
Treisman and Garry Gelade(1980)."A feature-integration theory of attention."
Cognitive Psychology,Vol.12,No.1,pp.97–136).Herein on basis, Itti and Koch et al. are proposed
Earliest conspicuousness detects computation model (L.Itti, C.Koch, E.Niebur, " Amodel of saliency-based
visual attention for rapid scene analysis”,IEEE Trans.Pattern
Anal.Mach.Intell 20 (11) (1998) 1254-1259), i.e., famous IT model.The conspicuousness detection algorithm of early stage is infused
Focus is watched in prediction attentively again, can not consistently protrude foreground target region, and includes a large amount of back in the notable figure formed
Scape noise, these problems significantly limit the application of conspicuousness detection algorithm.
With the continuous development of computer vision, in the especially close more than ten years, scholars propose a large amount of conspicuousness inspection
Method of determining and calculating, main thought are still the prominent foreground target by way of Characteristic Contrast.Based on the significant of local feature comparison
Property detection algorithm use center-periphery Comparing method, that is, pass through the Characteristic Contrast of middle section and its peripheral neighborhood, protrusion mesh
Mark object.This method often highlights the fringe region of foreground target and cannot consistently protrude entire foreground target.Based on complete
The conspicuousness detection algorithm of office's Characteristic Contrast often chooses suitable background information, for example, the feature in framing mask region, and
By comparing prominent foreground target.This method only considered the feature of selected background area, directly be regarded as background characteristics
For Characteristic Contrast, and the spatial distribution of background area feature is ignored, may include portion in extracted background characteristics therefore
Divide foreground information, negative effect will cause to the detection of subsequent conspicuousness.Therefore, the method based on Characteristic Contrast is capable of detecting when
Well-marked target under simple scenario, but for prospect or background there are many feature and the complex scene deposited, detection effect is still
It is so undesirable.
In recent years, scholars gradually recognize that the cognitive sciences such as Cognitive Neuroscience and cognitive psychology examine conspicuousness
The heuristic effect surveyed.For example, Wei et al. (Geodesic saliency using background priors) finder
The frame portion of image is often defaulted as background component when observing image by eye, thus introduces background priori, and pass through feature
Comparison form global conspicuousness detection.This method only used the prominent foreground area of feature of background component, not examine but
Consider the spatial distribution of its feature.To effectively inhibit the ambient noise in notable figure, the river the Lu Hu team of Dalian University of Technology
(J.Wang,H.Lu,X.Li,N.Tong,W.Liu,Saliency detection via background and
Foreground seed selection, Neurocomputing 152 (2015) 359-368.) in conspicuousness detection process
Foreground information is introduced, the convex hull for successively forming angle point in image and the area that adaptivenon-uniform sampling generation is carried out to background notable figure
Domain is considered as foreground area.But the convex hull of angle point formation does not account for the profile information of target object, and adaptivenon-uniform sampling does not have
Consider the compact nature of target object, therefore, the foreground information itself introduced using the above method can be made an uproar comprising a large amount of background
Sound, so that subsequent noise suppression effect is bad.
Summary of the invention
It is an object of the invention to overcome deficiency in the prior art, a kind of coloured silk based on background and foreground information is provided
Color image significance detection method, solving can not accomplish consistently to protrude in the prior art foreground target and effectively inhibit aobvious
The technical issues of writing the ambient noise in figure.
In order to solve the above technical problems, the present invention provides a kind of color image conspicuousness based on background and foreground information
Detection method, characterized in that the following steps are included:
Step 1, image preprocessing: carrying out over-segmentation to the color image of input and handle to obtain a series of super-pixel block, will
Super-pixel block is as minimal processing unit；
The conspicuousness detection based on background information: step 2 chooses background seed, passes through each super-pixel block and background seed
Between Characteristic Contrast obtain coarse significance；Feature distribution based on background seed defines the background weight of each super-pixel block, leads to
Crossing background weight improves coarse significance of the significance acquisition based on background information；
Step 3, based on foreground information conspicuousness detection: to previous step formed based on the notable figure of background information into
Row segmentation, chooses a close foreground area in all segmentation results, extracts foreground area feature, is obtained by Characteristic Contrast
Obtain the significance based on foreground information；
The integration of significance: step 4 it is whole to integrate obtaining based on the significance of background and foreground information for first two steps acquisition
Significance is closed, and carries out the significance after smooth operation obtains all super-pixel block optimizations to significance is integrated.
Further, in said step 1, over-segmentation processing uses SLIC superpixel segmentation method.
Further, in the step 2, the detailed process of the significance based on background information is obtained are as follows:
11) super-pixel block at framing mask is chosen as background seed, passes through super-pixel block each in image and background kind
Son carries out the coarse significance that Characteristic Contrast obtains each super-pixel block；
12) K mean cluster is carried out to selected background seed, determines that each cluster belongs to according to the spatial distribution of each cluster
The background weight of background seed is defined as follows in the probability of background, k-th of cluster:
Pk=1-exp (- α (Ls+Lo)) (k=1,2 ..., K)
Wherein, LsLength for the most short super-pixel chain clustered comprising all k-th, LoTo belong to it in the super-pixel chain
The quantity for the super-pixel block that he clusters, parameter alpha range are that 0.01~0.08, K is the cluster centre number chosen；
13) for other super-pixel block in image, firstly, the geodesic distance of super-pixel block and had powerful connections seed is calculated,
It obtains and the smallest background seed of the super-pixel block geodesic distance:
Wherein, BG is the set of background seed, dgeo(si,sj) be two super-pixel block geodesic distance；By previous step 12)
The background probability for knowing the background seed, remembers that the background probability of the background seed isThis super-pixel block and this background seed
Geodesic distance isThe then background weight of this super-pixel block are as follows:
Then the background weight of each super-pixel block is successively calculated；
14) significance based on background information of super-pixel block is defined are as follows:
Wherein,For super-pixel block siThe significance based on background information,For the super picture being calculated in step 11)
Plain block siCoarse significance.
Further, parameter alpha 0.05.
Further, in the step 3, the significance detailed process based on foreground information is obtained are as follows:
21) notable figure based on background obtained in previous step is divided using parametric maxflow method
It cuts, obtains a series of close foreground areas, maxflow method segmentation result are as follows:
Wherein, N is super-pixel number in image, AiFor super-pixel block siArea, xi∈ { 1,0 } indicates super-pixel block si
Whether foreground area, e are belonged toijFor the similarity between neighbouring super pixels block, xfFor obtained foreground area segmentation result；
22) in all segmentation results, the segmentation result optimal according to following formula selected value is as foreground area:
Wherein, XfTo divide obtained multiple segmentation results, N is the number of super-pixel block,For super-pixel block siBase
In the significance of background, V (xf) it is segmentation result xfSpace coordinate variance；
23) the foreground area feature for extracting selection, by Characteristic Contrast determine each super-pixel block in image based on prospect
Significance:
Wherein, FG is the set of the prospect super-pixel block obtained, dc(si,sj) and dl(si,sj) be respectively super-pixel block it
Between color distance and space length.
Further, in the step 4, detailed process are as follows:
31) significance based on background and prospect of acquisition is integrated, it is as follows integrates formula:
Wherein,For super-pixel block siIntegration significance,Indicate super-pixel block siThe significance based on background,Indicate super-pixel block siThe significance based on prospect, parameter P value range be 2.5~8,
32) significance of integration is advanced optimized, majorized function is as follows:
Wherein, wc(si,sj) indicate the color similarities of adjacent two super-pixel block, SiAnd SjIndicate to be optimized adjacent
The significance of two super-pixel block,WithRespectively super-pixel block siBackground weight and prospect label,For super-pixel block
siIntegration significance, N be image in super-pixel number；This majorized function is global optimization function, is carried out to majorized function
Optimization solves the optimization significance of all super-pixel block.
Further, parameter P value 4.
Compared with prior art, the beneficial effects obtained by the present invention are as follows being: the present invention uses special based on framing mask region
The background weight for levying distribution improves the conspicuousness detection effect based on background characteristics comparison；Before being extracted based on maxflow method
Scene area has taken into account the marginal information of foreground target and the compactness of target object, being capable of prospect mesh in accurate description scene
Mark；Two notable figures that the different role of background and foreground information in conspicuousness detection is integrated, advanced optimize integration
The notable figure of obtained notable figure, optimization is more smooth-out inside background and foreground area.The present invention can be more consistent
Foreground target in the prominent image in ground, and have good inhibitory effect to ambient noise in image.
Detailed description of the invention
Fig. 1 is the flow diagram that the color image conspicuousness of the embodiment of the present invention detects.
Fig. 2 is that the background weight of the embodiment of the present invention improves the effect picture of conspicuousness detection, wherein (a) is input picture,
(b) it is well-marked target true value, (c) is coarse notable figure, (d) be background weight, (e) is the notable figure based on background.
Fig. 3 is the background weight schematic diagram of calculation flow of the embodiment of the present invention, wherein (a) is input picture, is (b) super
Pixel segmentation result, (c) is background seed cluster result, (d) is the background weight of selected seed, (e) is all super-pixel block
Background weight.
Fig. 4 is that the effect picture of noise is extracted and its inhibited to the foreground area of the embodiment of the present invention, wherein (a) is input figure
Picture (b) is the notable figure based on background, is (c) foreground area extracted, (d) is the notable figure based on prospect, is (e) integration
Notable figure, (f) for optimization notable figure.
Fig. 5 is the comparison diagram of the conspicuousness testing result of the embodiment of the present invention and the testing result of the prior art, wherein
(a) it is input picture, (b) is well-marked target true value, (c) be the testing result of the present patent application, (d) is the detection knot of IT model
Fruit is (e) testing result of XIE model, (f) is the testing result of BFS model.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.Following embodiment is only used for clearly illustrating the present invention
Technical solution, and not intended to limit the protection scope of the present invention.
Fig. 1 is the flow diagram that the color image conspicuousness of the embodiment of the present invention detects.As shown in Figure 1, this method
A kind of color image conspicuousness detection method based on background and foreground information, characterized in that the following steps are included:
Step 1, image preprocessing: carrying out over-segmentation to the color image of input and handle to obtain a series of super-pixel block, will
Super-pixel block is as minimal processing unit.
The color image of input is utilized into SLIC superpixel segmentation method, image is excessively segmented into many super-pixel block,
Using super-pixel block as the smallest processing unit of subsequent operation.
Step 2, the conspicuousness detection based on background information:
Human eye tends to pay close attention to when observing image the center (it has been generally acknowledged that center that target appears in image) of image,
And ignore the frame region of image (frame of image is background).Therefore the super-pixel block at framing mask is chosen as background kind
Son carries out the coarse significance that Characteristic Contrast obtains all super-pixel block, shape by super-pixel block each in image and background seed
At coarse notable figure, as shown in Fig. 2 (c).In order to exclude the interference of foreground information, the spy of selected background seed is further considered
Sign distribution carries out feature clustering to background seed, determines that seed belongs to the probability of background according to the spatial distribution of each cluster, according to
This defines the background weight of each super-pixel block, as shown in Fig. 2 (d).Improve coarse significance with background weight, obtains based on back
The significance of scape information forms the notable figure based on background, as shown in Fig. 2 (e).
Obtain the detailed process of the notable figure based on background information are as follows:
11) super-pixel block at framing mask is chosen as background seed, passes through super-pixel block each in image and background kind
Son carries out the coarse significance that Characteristic Contrast obtains each super-pixel block, forms coarse notable figure, this process refers to the prior art；
12) K mean cluster is carried out to selected background seed, as shown in Fig. 3 (c), according to the space of each cluster point
Cloth determines each probability for clustering and belonging to background, and the background weight of background seed is defined as follows in k-th of cluster:
Pk=1-exp (- α (Ls+Lo)) (k=1,2 ..., K)
Wherein, LsLength for the most short super-pixel chain clustered comprising all k-th, LoTo belong to it in the super-pixel chain
The quantity for the super-pixel block that he clusters, parameter alpha is constant, can be set to 0.01~0.08, determines that α is arranged by actual tests
It can get optimum detection effect when being 0.05, K is the cluster centre number chosen；The background weight value of super-pixel block is bigger, belongs to
It is bigger in the probability of background, on the contrary, value is smaller, then it is assumed that it is bigger that it belongs to a possibility that prospect.
13) for other super-pixel block in image, it is calculated according to the connectivity of super-pixel block and selected background seed
Background weight.Firstly, calculating the geodesic distance of super-pixel block and had powerful connections seed, obtain with the super-pixel block geodesic distance most
Small background seed:
Wherein, BG is the set of background seed, dgeo(si,sj) be two super-pixel block geodesic distance.By previous step 12)
The background probability for knowing the background seed, remembers that the background probability of the background seed isThis super-pixel block and this background seed
Geodesic distance isThe then background weight of this super-pixel block are as follows:
It calculates shown in effect such as Fig. 3 (e)；
14) significance based on background information of super-pixel block is defined are as follows:
Wherein,For super-pixel block siThe significance based on background information,It is super for what is be calculated in step 11)
Block of pixels siCoarse significance, the significance by each super-pixel block based on background information obtains based on the significant of background information
Figure, as shown in Fig. 2 (e).
Step 3, the conspicuousness detection based on foreground information: the notable figure based on background obtained in previous step carries out
A close foreground area is chosen in segmentation in all segmentation results, extracts foreground target feature, is obtained by Characteristic Contrast
Significance based on foreground information forms the notable figure based on foreground information.
Obtain the detailed process of the notable figure based on foreground information are as follows:
21) significance based on background obtained in previous step is divided using parametric maxflow method
It cuts, obtains a series of close foreground areas:
Wherein, N is super-pixel number in image, AiFor super-pixel block siArea, xi∈ { 1,0 } indicates super-pixel block si
Whether foreground area, e are belonged toijFor the similarity between neighbouring super pixels block, xfFor obtained foreground area segmentation result.It compares
In OTSU method, the foreground area that this method is divided is consistent with the characteristic of the tight close of well-marked target, can be more preferable
Ground describes the well-marked target in image, as shown in Fig. 4 (c)；
22) according to each segmentation result and the consistent degree of the notable figure based on background and the space precise of conspicuousness target
Property, most suitable segmentation result is chosen as foreground area according to following formula:
Wherein, XfTo divide obtained multiple segmentation results, N is the number of super-pixel block,For super-pixel block siBase
In the significance of background, V (xf) it is segmentation result xfSpace coordinate variance.As shown in Fig. 4 (c), this foreground area extracting method
Both considered the marginal information of foreground target it is contemplated that target object compactness, extracted foreground area can be accurately anti-
Reflect foreground target feature；
23) extract foreground area feature, by Characteristic Contrast determine each super-pixel block in image based on the significant of prospect
Degree:
Wherein, FG is the set of the prospect super-pixel block obtained, dc(si,sj) and dl(si,sj) be respectively super-pixel block it
Between color distance and space length.Notable figure based on prospect is obtained by the significance based on prospect of each super-pixel block, such as
Shown in Fig. 4 (d), since background area and foreground area often have color and difference spatially, calculated using the method
To the notable figure based on prospect can effectively inhibit ambient noise.
Step 4, the integration and optimization of notable figure:
Since the effect of background information and foreground information in conspicuousness detection is different, i.e., background information is for protruding prospect
Target, and foreground information integrates the notable figure based on background and foreground information of first two steps acquisition for inhibiting ambient noise.Base
It is as follows that formula is integrated in the significance of background and prospect:
Wherein,For the integration significance of i-th of super-pixel block,Indicate the aobvious based on background of i-th of super-pixel block
Work degree,Indicate the significance based on prospect of i-th of super-pixel block, parameter P value is constant, parameter area can for 2.5~
8, it can get optimum detection effect when determining that β is set as 4 by actual tests；Prospect can either be protruded by integrating obtained notable figure
Target restrained effectively ambient noise again.
Visually more smooth notable figure in order to obtain advanced optimizes the saliency map of integration, and majorized function is as follows:
Wherein, wc(si,sj) indicate the color similarities of adjacent two super-pixel block, SiAnd SjIndicate to be optimized adjacent
The significance of two super-pixel block,WithRespectively super-pixel block siBackground weight and prospect label,For super-pixel block
siIntegration significance, N be image in super-pixel number.This majorized function is global optimization function, is carried out to majorized function
It solves, solution procedure is formed final referring to the prior art, the last optimization significance for directly once solving all super-pixel block
Notable figure based on background and foreground information.
Fig. 5 is the comparison diagram according to the conspicuousness testing result of the embodiment of the present invention and the testing result of the prior art.
Wherein, Fig. 5 (c) is the testing result of the present patent application, and Fig. 5 (d) is IT model (L.Itti, C.Koch, E.Niebur, A
model of saliency-based visual attention for rapid scene analysis,IEEE
Trans.Pattern Anal.Mach.Intell 20 (11) (1998) 1254-1259.) testing result, Fig. 5 (e) be XIE
Model (Y.Xie, H.Lu, M.-H.Yang, Bayesian saliency via low and mid level cues, IEEE
Trans.Image Processing 22 (5) (2013) 1689-1698.) testing result, Fig. 5 (f) be BFS model
(J.Wang,H.Lu,X.Li,N.Tong,W.Liu,Saliency detection via background and
Foreground seed selection, Neurocomputing 152 (2015) 359-368.) testing result.IT model
It is that one kind watches Focus prediction model attentively, can not consistently protrudes entire well-marked target.XIE model and BFS model pass through respectively
The region that angle point convex hull and adaptivenon-uniform sampling generate introduces foreground information, usually contains background parts in the region generated, no
It can accurately reflect foreground target feature.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, several improvements and modifications, these improvements and modifications can also be made
Also it should be regarded as protection scope of the present invention.
Claims (6)
1. a kind of color image conspicuousness detection method based on background and foreground information, characterized in that the following steps are included:
Image preprocessing: step 1 carries out over-segmentation to the color image of input and handles to obtain a series of super-pixel block, by super picture
Plain block is as minimal processing unit；
The conspicuousness detection based on background information: step 2 chooses background seed, by between each super-pixel block and background seed
Characteristic Contrast obtains coarse significance；Feature distribution based on background seed defines the background weight of each super-pixel block, passes through back
Scape weight improves coarse significance and obtains the significance based on background information；
Step 3, the conspicuousness detection based on foreground information: the notable figure based on background information formed to previous step is divided
It cuts, a close foreground area is chosen in all segmentation results, extract foreground area feature, base is obtained by Characteristic Contrast
In the significance of foreground information；
The integration of significance: step 4 integrates the aobvious based on the acquisition integration of the significance of background and foreground information of first two steps acquisition
Work degree, and the significance after smooth operation obtains all super-pixel block optimizations is carried out to significance is integrated；
In the step 2, the detailed process of the significance based on background information is obtained are as follows:
11) choose the super-pixel block at framing mask and be used as background seed, by super-pixel block each in image and background seed into
Row Characteristic Contrast obtains the coarse significance of each super-pixel block；
12) K mean cluster is carried out to selected background seed, determines that each cluster belongs to back according to the spatial distribution of each cluster
The probability of scape, the background weight of background seed is defined as follows in k-th of cluster:
Pk=1-exp (- α (Ls+Lo)), k=1,2 ..., K
Wherein, LsLength for the most short super-pixel chain clustered comprising all k-th, LoIt is poly- to belong to other in the super-pixel chain
The quantity of the super-pixel block of class, parameter alpha range are that 0.01~0.08, K is the cluster centre number chosen；
13) it for other super-pixel block in image, firstly, calculating the geodesic distance of super-pixel block and had powerful connections seed, obtains
With the smallest background seed of the super-pixel block geodesic distance:
Wherein, BG is the set of background seed, dgeo(si,sj) be two super-pixel block geodesic distance；From previous step 12)
The background probability of the background seed remembers that the background probability of the background seed isThe geodetic of this super-pixel block and this background seed
Distance isThe then background weight of this super-pixel block are as follows:
Then the background weight of each super-pixel block is successively calculated；
14) significance based on background information of super-pixel block is defined are as follows:
Wherein,For super-pixel block siThe significance based on background information,For the super-pixel block being calculated in step 11)
siCoarse significance.
2. a kind of color image conspicuousness detection method based on background and foreground information according to claim 1, special
Sign is that in said step 1, over-segmentation processing uses SLIC superpixel segmentation method.
3. a kind of color image conspicuousness detection method based on background and foreground information according to claim 1, special
Sign is parameter alpha 0.05.
4. a kind of color image conspicuousness detection method based on background and foreground information according to claim 1, special
Sign is, in the step 3, obtains the significance detailed process based on foreground information are as follows:
21) notable figure based on background obtained in previous step is split using parametric maxflow method,
Obtain a series of close foreground areas, maxflow method segmentation result are as follows:
Wherein, N is super-pixel number in image, AiFor super-pixel block siArea, xi∈ { 1,0 } indicates super-pixel block siWhether belong to
In foreground area, eijFor the similarity between neighbouring super pixels block, xfFor obtained foreground area segmentation result；
22) in all segmentation results, the segmentation result optimal according to following formula selected value is as foreground area:
Wherein, xfTo divide obtained multiple segmentation results, N is the number of super-pixel block,For super-pixel block siBased on background
Significance, V (xf) it is segmentation result xfSpace coordinate variance；
23) the foreground area feature for extracting selection determines the showing based on prospect of each super-pixel block in image by Characteristic Contrast
Work degree:
Wherein, FG is the set of the prospect super-pixel block obtained, dc(si,sj) and dl(si,sj) it is respectively between super-pixel block
Color distance and space length.
5. a kind of color image conspicuousness detection method based on background and foreground information according to claim 4, special
Sign is, in the step 4, detailed process are as follows:
31) significance based on background and prospect of acquisition is integrated, it is as follows integrates formula:
Wherein,For super-pixel block siIntegration significance,Indicate super-pixel block siThe significance based on background,It indicates
Super-pixel block siThe significance based on prospect, parameter P value range be 2.5~8；
32) significance of integration is advanced optimized, majorized function is as follows:
Wherein, wc(si,sj) indicate the color similarities of adjacent two super-pixel block, SiAnd SjIndicate to be optimized two neighboring
The significance of super-pixel block,WithRespectively super-pixel block siBackground weight and prospect label,For super-pixel block si's
Significance is integrated, N is the number of super-pixel in image；This majorized function is global optimization function, is optimized to majorized function
Solve the optimization significance of all super-pixel block.
6. a kind of color image conspicuousness detection method based on background and foreground information according to claim 5, special
Sign is parameter P value 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610654316.1A CN106327507B (en) | 2016-08-10 | 2016-08-10 | A kind of color image conspicuousness detection method based on background and foreground information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610654316.1A CN106327507B (en) | 2016-08-10 | 2016-08-10 | A kind of color image conspicuousness detection method based on background and foreground information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106327507A CN106327507A (en) | 2017-01-11 |
| CN106327507B true CN106327507B (en) | 2019-02-22 |
Family
ID=57740141
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610654316.1A Expired - Fee Related CN106327507B (en) | 2016-08-10 | 2016-08-10 | A kind of color image conspicuousness detection method based on background and foreground information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106327507B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106709474A (en) * | 2017-01-23 | 2017-05-24 | 无锡职业技术学院 | Handwritten telephone number identification, verification and information sending system |
| CN107016682B (en) * | 2017-04-11 | 2020-03-31 | 四川大学 | Self-adaptive segmentation method for salient objects of natural images |
| CN107194870B (en) * | 2017-05-24 | 2020-07-28 | 北京大学深圳研究生院 | Image scene reconstruction method based on salient object detection |
| CN107452013A (en) * | 2017-05-27 | 2017-12-08 | 深圳市美好幸福生活安全系统有限公司 | Conspicuousness detection method based on Harris Corner Detections and Sugeno fuzzy integrals |
| CN108198172B (en) * | 2017-12-28 | 2022-01-28 | 北京大学深圳研究生院 | Image significance detection method and device |
| CN108965739A (en) * | 2018-06-22 | 2018-12-07 | 北京华捷艾米科技有限公司 | video keying method and machine readable storage medium |
| CN109166106B (en) * | 2018-08-02 | 2021-07-30 | 山东大学 | Target detection position correction method and device based on sliding window |
| CN109472259B (en) * | 2018-10-30 | 2021-03-26 | 河北工业大学 | Image collaborative saliency detection method based on energy optimization |
| CN110310263B (en) * | 2019-06-24 | 2020-12-01 | 北京师范大学 | SAR image residential area detection method based on significance analysis and background prior |
| CN110866896B (en) * | 2019-10-29 | 2022-06-24 | 中国地质大学（武汉） | Image saliency target detection method based on k-means and level set super-pixel segmentation |
| CN112183556B (en) * | 2020-09-27 | 2022-08-30 | 长光卫星技术股份有限公司 | Port ore heap contour extraction method based on spatial clustering and watershed transformation |
| CN113378873A (en) * | 2021-01-13 | 2021-09-10 | 杭州小创科技有限公司 | Algorithm for determining attribution or classification of target object |
| CN112861858A (en) * | 2021-02-19 | 2021-05-28 | 首都师范大学 | Significance truth diagram generation method and significance detection model training method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513070A (en) * | 2015-12-07 | 2016-04-20 | 天津大学 | RGB-D salient object detection method based on foreground and background optimization |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2523329A (en) * | 2014-02-20 | 2015-08-26 | Nokia Technologies Oy | Method, apparatus and computer program product for image segmentation |
-
2016
- 2016-08-10 CN CN201610654316.1A patent/CN106327507B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513070A (en) * | 2015-12-07 | 2016-04-20 | 天津大学 | RGB-D salient object detection method based on foreground and background optimization |
Non-Patent Citations (3)
| Title |
|---|
| A Model of Saliency-Based Visual Attention for Rapid Scene Analysis;Laurent Itti, et al.;《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》;19981130;第20卷(第11期);第1254-1259页 |
| Saliency detection via background and foreground seed selection;Jianpeng Wang, et al.;《NEUROCOMPUTING》;20151231;第152卷;第359页摘要，第360-363页第2节 |
| 基于高斯超像素的快速Graph Cuts图像分割方法;韩守东 等;《自动化学报》;20110131;第37卷(第1期);第11-20页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106327507A (en) | 2017-01-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106327507B (en) | A kind of color image conspicuousness detection method based on background and foreground information | |
| CN107730515B (en) | Increase the panoramic picture conspicuousness detection method with eye movement model based on region | |
| Khan | Image segmentation techniques: A survey | |
| CN106250895B (en) | A kind of remote sensing image region of interest area detecting method | |
| Sun et al. | Recognition of green apples in an orchard environment by combining the GrabCut model and Ncut algorithm | |
| Morris | A pyramid CNN for dense-leaves segmentation | |
| CN105279769B (en) | A kind of level particle filter tracking method for combining multiple features | |
| CN108664838A (en) | Based on the monitoring scene pedestrian detection method end to end for improving RPN depth networks | |
| CN109886170A (en) | A kind of identification of oncomelania intelligent measurement and statistical system | |
| CN110533100A (en) | A method of CME detection and tracking is carried out based on machine learning | |
| Song et al. | Feature extraction and target recognition of moving image sequences | |
| Li et al. | Domain adaptive box-supervised instance segmentation network for mitosis detection | |
| Liu et al. | Double mask R-CNN for pedestrian detection in a crowd | |
| Yan et al. | Robust infrared superpixel image separation model for small target detection | |
| Hu et al. | Fast face detection based on skin color segmentation using single chrominance Cr | |
| Sohail et al. | Deep object detection based mitosis analysis in breast cancer histopathological images | |
| Zhang et al. | An Improved Computational Approach for Salient Region Detection. | |
| Liang et al. | Human-guided flood mapping: From experts to the crowd | |
| Li et al. | Fingertip blood collection point localization research based on infrared finger vein image segmentation | |
| CN113012167B (en) | Combined segmentation method for cell nucleus and cytoplasm | |
| Pan et al. | Preferential image segmentation using trees of shapes | |
| Zhang et al. | Land Cover Change Detection Based on Vector Polygons and Deep Learning With High Resolution Remote Sensing Images | |
| CN110751034B (en) | Pedestrian behavior recognition method and terminal equipment | |
| Fang et al. | Recognizing global dams from high-resolution remotely sensed images using convolutional neural networks | |
| Zhang et al. | Saliency and background prior-based residential area detection for SAR images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190222 Termination date: 20210810 |