CN108694710A - One kind being based on the notable figure fusion method of (N) fuzzy integral - Google Patents
One kind being based on the notable figure fusion method of (N) fuzzy integral Download PDFInfo
- Publication number
- CN108694710A CN108694710A CN201810375979.9A CN201810375979A CN108694710A CN 108694710 A CN108694710 A CN 108694710A CN 201810375979 A CN201810375979 A CN 201810375979A CN 108694710 A CN108694710 A CN 108694710A
- Authority
- CN
- China
- Prior art keywords
- notable
- formula
- width
- fuzzy
- value
- 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
- 238000007500 overflow downdraw method Methods 0.000 title claims abstract description 10
- 238000001514 detection method Methods 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 19
- 230000004927 fusion Effects 0.000 claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims abstract description 12
- 238000004364 calculation method Methods 0.000 claims description 6
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 10
- 238000012360 testing method Methods 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 3
- 238000005457 optimization Methods 0.000 abstract description 2
- 230000002708 enhancing effect Effects 0.000 abstract 1
- 238000010189 synthetic method Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 8
- 230000007812 deficiency Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 101001013832 Homo sapiens Mitochondrial peptide methionine sulfoxide reductase Proteins 0.000 description 1
- 102100031767 Mitochondrial peptide methionine sulfoxide reductase Human genes 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
- G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to technical field of image processing, are related to a kind of notable figure fusion method being based on (N) fuzzy integral.First, respective notable figure is generated using the conspicuousness detection method to be merged.Secondly, the similarity factor and similar matrix between each notable figure are calculated, and then obtain the degree of being supported and confidence level of each width notable figure.Then, using the confidence level of each notable figure as the fuzzy mearue value in (N) fuzzy integral.At the same time, the sequence that Pixel-level is carried out to the notable figure to be merged, using the discrete saliency value of sequence as the non-negative real value measurable function in (N) fuzzy integral.Then, calculate the notable figure that (N) fuzzy integral is worth to after fusion inhibits the optimization processing of background area to obtain final notable figure finally by enhancing foreground area.This method can recognize that the highlight in image, merge the existing outstanding respective advantage of conspicuousness detection method, and obtained detection result carries out alone effect when significance test better than each synthetic method.
Description
Technical field
The invention belongs to technical field of image processing, are related to a kind of notable figure fusion method being based on (N) fuzzy integral.
Background technology
Saliency detection is intended to find out most important part in image, is that computer vision field is used for reducing calculating
The important pre-treatment step of complexity has a wide range of applications, together in fields such as compression of images, target identification, image segmentations
When it is problem challenging in computer vision again, attract the research interest of a large amount of scholars.There has been big
The method for measuring outstanding saliency detection, these methods each have the advantage of oneself and deficiency, even same aobvious
Work property detection method, is also that difference is huge for different picture detection results.A variety of conspicuousness detections can be merged thus
Method as a result, being just particularly important in the method for obtaining more excellent notable figure.The side for the notable figure fusion for having some traditional
Method, it is simply to sum it up average or simple multiplication for the progress of several notable figures to be averaged that they are mostly, and this notable figure merges
Mode puts on an equal footing various notable figures, the weights of various conspicuousness detection methods is set as same numerical value, this is in practice
Unreasonable, because for a width picture even each pixel, the detection result of various conspicuousness detection methods is all
Different, difference ought to be also arranged in the weights of each conspicuousness detection method thus.Currently there is also some research fusions, several are aobvious
The method for writing figure, as Mai et al. merges several notable figures using condition random field (CRF), but calculating speed is too slow;Qin etc.
People merges several notable figures using multilayer cellular automata (MCA), has obtained extraordinary effect, but in terms of its recall rate
Effect is not satisfactory.
Present invention discover that the superiority that (N) fuzzy integral is shown in multiple Classifiers Combination, it is contemplated that (N) fuzzy integral
In fuzzy mearue the significance level of various decisions is taken into account, merging a variety of conspicuousness detection algorithms, to obtain fusion aobvious for this
It is very rational when work figure, but (N) fuzzy integral is applied not yet in conspicuousness detection field, (N) is obscured product by the present invention
Point application range expands to saliency detection field, uses (N) fuzzy integral to carry out Pixel-level to several notable figures and melts
The advantages of closing, making full use of multiple notable figures improves the effect of conspicuousness detection, with obvious effects independent higher than each after fusion
Conspicuousness detection method effect, obtained good effect in accuracy rate, recall rate, F-measure values.
Invention content
The present invention proposes a kind of notable figure fusion method being based on (N) fuzzy integral, it is therefore intended that overcomes above-mentioned existing
The deficiency of technology obtains a kind of fusion method of more preferably several notable figures.
Technical scheme of the present invention:
One kind being based on the notable figure fusion method of (N) fuzzy integral, and steps are as follows:
The first step generates n width notable figures using the n kind methods to be merged, and n > 1 are combined to obtain limited by n width notable figures
Set X={ x1, x2..., xn, wherein xnIndicate the n-th width notable figure;Each width notable figure is all considered as a kind of grader, it will
Conspicuousness detection is considered as a kind of two classification problems, and the matrix D P for obtaining decision section is indicated as shown in formula (1):
Wherein first is classified as each grader judgement pixel as the possibility of foreground, and second is classified as each grader judgement pixel
Point is the possibility of background, and a kind of grader is represented per a line;I-th of grader judges pixel for the probability F of foregroundiIt indicates
For Fi=f (xi), 1≤i≤n judges pixel for the probability B of backgroundiIt is expressed as Bi=1-f (xi), wherein f (xi) it is the i-th width
The saliency value of notable figure, Pixel-level merge notable figure, and the saliency value of notable figure refers both to correspond to notable at a certain pixel
Value;It is converted into for this purpose, notable figure fusion method will merge the problem of several notable figures and asks DP matrix first rows corresponding (N) fuzzy
The problem of integral;
Second step calculates related coefficient between each width notable figure, finds out similar matrix;Similarity factor dijCalculating such as formula
(2) shown in:
Wherein , ||f(xi), f (xj)||It indicates in CIELab color spaces, the i-th width notable figure is aobvious with jth width notable figure
The Euclidean distance of work value, σ2=0.1;Similarity factor dijFor describing the similarity degree between each width notable figure, dij∈[0,1],
The bigger expression notable figure x of valueiWith xjBetween it is more similar;
By related coefficient, obtain shown in the similar matrix such as formula (3) corresponding to n width notable figures:
Third walks, and finds out the support and confidence level between each width notable figure;Notable figure xiDegree of being supported indicate by other
The degree of support of notable figure, if a width notable figure and other notable figures are all more similar, then it is assumed that their mutual support degree
Also higher, notable figure xiDegree of being supported Sup (xi) shown in calculation formula such as formula (4):
The confidence level of notable figure reflects that certain width notable figure obtains the credibility of saliency value, and a width notable figure is notable by other
The supported degree of figure is higher, and the notable figure confidence level is bigger, and the confidence level of notable figure is calculated as shown in formula (5):
4th step finds out the fuzzy mearue of each width notable figure composition finite aggregate X power sets;Defining Y is made of the subset of X
Power set, μ:Y→[0,1]It indicates from power set Y to [0,1]Mapping, take μ be fuzzy mearue, meet μ (φ)=0, μ (X)=
1, it is canonical;Shown in λ fuzzy mearues calculation formula such as formula (6):
μ (A ∪ B)=μ (A)+μ (B)+λ μ (A) μ (B) (6)
Wherein,A ∩ B=φ, there are constant λ > -1;The value of λ by solution formula (7) and formula (8) come
It arrives;
μi=μ ({ xi)=Crd (xi) (8)
Wherein, μiIndicate notable figure xiTo the significance level of final fusion results, other measure values in power set Y are by formula
(6) it is calculated;
5th step defines the non-negative real value measurable function f in (N) fuzzy integral at each pixel of notable figure:X→
[0,1]It is discrete-valued function, functional value is { f (x1), f (x2) ..., f (xn), wherein f (xn) it is the notable of the n-th width notable figure
Value;N width notable figures are ranked up so that the function f after sequence meets 0≤f (xθ(0))≤f(xθ(1))≤…≤f(xθ(n))≤
1, wherein θ is ranking functions;
6th step calculates the notable figure that (N) fuzzy integral is merged, definition such as formula (9) institute of (N) fuzzy integral
Show:
Wherein, meet ∨ representatives to be maximized, formula (10) and formula (11) are further obtained by formula (9):
Aθ(i)={ xθ(i), xθ(i+1)..., xθ(n), i=1,2 ..., n (11)
Wherein, μ (xθ(i)) be and f (xθ(i)) corresponding fuzzy mearue;The value that (N) ∫ f (x) d μ are calculated is at certain
At one pixel, (N) fuzzy integral is used to merge saliency value Sal (x) after the fusion that n kind notable figures obtain, such as formula (12) institute
Show:
Sal (x)=(N) ∫ f (x) d μ (12)
7th step enhances foreground area, inhibits background area;Obtained notable figure is optimized using formula (13)
Obtain final notable figure
Beneficial effects of the present invention are:This method is different from existing methods characteristic and is to comprehensively utilize various conspicuousness inspections
The advantages of survey method, result of the obtained effect better than each individually conspicuousness detection method.It is shown with traditional merging simultaneously more
The method of work figure is compared, and the present invention treats the various conspicuousness detection methods to be merged with a certain discrimination, assigns different weights respectively, effect
Fruit is more excellent.In addition (N) fuzzy integrals theory has also been introduced into conspicuousness detection field by the present invention for the first time, and has been obtained either
Accuracy rate, recall rate are still averaged F-measure values all preferably effects.
Description of the drawings
Fig. 1 is the flow diagram of the method for the present invention;
Fig. 2 is the conspicuousness testing result comparison diagram of algorithms of different;(a) picture to be detected, (b) true value, (c) BSCA algorithms
Obtained significant result, (d) significant result that DSR algorithms obtain, (e) significant result that HS algorithms obtain, (f) wCO
The significant result that algorithm obtains, (g) significant result that MR algorithms obtain, (h) the conspicuousness testing result that the present invention obtains;
Fig. 3 is the present invention and PR (accuracy rate, recall rate) of other conspicuousness detection methods on ECSSD standard databases
Curve;
Fig. 4 is that the PR of the invention with other conspicuousness detection methods on DUT-OMRON standard databases (call together by accuracy rate
The rate of returning) curve;
Fig. 5 is that the PR of the invention with other conspicuousness detection methods on MSRA10K standard databases (recall by accuracy rate
Rate) curve;
Fig. 6 be the present invention with Average Accuracy of other conspicuousness detection methods on ECSSD standard databases, averagely call together
The block diagram of the rate of returning, average F-measure values;
Fig. 7 is the present invention and Average Accuracy of other conspicuousness detection methods on DUT-OMRON standard databases, flat
The block diagram of equal recall rate, average F-measure values;
Fig. 8 is the present invention and Average Accuracy of other conspicuousness detection methods on MSRA10K standard databases, average
The block diagram of recall rate, average F-measure values.
Specific implementation mode
Detailed description of embodiments of the present invention below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention
Premise is implemented, and gives detailed embodiment and specific operating process, but protection scope of the present invention is not limited to down
The embodiment stated.
The present invention tests the algorithm of proposition on the database of three standards:ECSSD databases, the database packet
1000 pictures are contained, picture size is different and there are many target, and some of pictures are derived from very difficult 300 data of Berkeley
Library.MSRA10K databases, it is the extension of MSRA databases, including 10000 pictures, cover all in ASD data sets
1000 pictures contain the picture of many complex backgrounds.DUT-OMRON databases include 5168 pictures in the database,
True value mark containing pixel scale, picture background is complicated, and target sizes are different, has prodigious challenge.These three data
There is the corresponding salient region figure manually demarcated in library.
Fig. 1 is the flow diagram of the method for the present invention;Fig. 2 is the conspicuousness testing result of the present invention and other algorithms of different
Comparison diagram;Fig. 3, Fig. 4, Fig. 5 are that different conspicuousness detection methods PR (accuracy rate, recall rate) on three standard databases is bent
Line;Fig. 6, Fig. 7, Fig. 8 are different conspicuousness detection methods Average Accuracies, average recall rate, flat on three standard databases
Equal F-measure value block diagrams.Realize the present invention the specific steps are:
The first step, using BSCA, DSR, HS, 5 kinds of conspicuousness detection methods such as wCO, MR generate the initial notable figure of 5 width.By
This 5 width notable figure combines to obtain finite aggregate X={ x1, x2, x3, x4, x5, wherein x1Indicate the notable figure that BSCA methods obtain,
x2Indicate the notable figure that DSR methods obtain, and so on.Conspicuousness detection is considered as a kind of two classification problems in the present invention, is obtained
Matrix D P to decision section is indicated as shown in formula (1):
Wherein first is classified as each grader judgement pixel as the possibility of foreground, and second is classified as each grader judgement pixel
Point is the possibility of background, and a kind of grader is represented per a line.A grader judgement pixel in i-th (1≤i≤5) is foreground
Probability FiIt is expressed as Fi=f (xi), judge pixel for the probability B of backgroundiIt is expressed as Bi=1-f (xi), wherein f (xi) it is i-th
(present invention is merged to notable figure in Pixel-level to the saliency value of width notable figure, the saliency value for the notable figure pointed out in invention
It refers both to correspond to the saliency value at a certain pixel).The problem of several notable figures will be merged in the present invention thus, which is converted into, seeks DP matrixes
Corresponding (N) the fuzzy integral problem of first row.
Second step calculates related coefficient between 5 width notable figures, finds out similar matrix.Similarity factor dijCalculating such as formula (2)
It is shown:
Qi Zhong ||f(xi), f (xj)||Indicate that i-th (1≤i≤5) width notable figure and jth (1≤j≤5) width notable figure are notable
The Euclidean distance of value, the present invention in σ2=0.1.Similarity factor dijFor describing the similarity degree between notable figure, dij∈[0,1],
The bigger expression notable figure x of its valueiWith xjBetween it is more similar.
By related coefficient, can be obtained in the present invention shown in the similar matrix such as formula (3) corresponding to 5 width notable figures:
Third walks, and finds out the support and confidence level between each notable figure.Notable figure xiDegree of being supported indicate by other 4 width
The degree that notable figure is supported, if a width notable figure and other 4 width notable figures are all more similar, then it is assumed that their mutual support
Spend also higher, notable figure xiDegree of being supported Sup (xi) shown in calculation formula such as formula (4):
The confidence level of notable figure reflects certain width notable figure and obtains the credibility of saliency value, and general width notable figure is by it
The degree that his other 4 width notable figures are supported is higher, and the notable figure confidence level is bigger, and the confidence level of notable figure calculates such as formula
(5) shown in:
4th step finds out the fuzzy mearue of each notable figure composition finite aggregate X power sets.Y defined in the present invention is the son by X
Collect the power set constituted, μ:Y→[0,1]It indicates from power set Y to [0,1]Mapping, the present invention in take μ be fuzzy mearue, meet μ
(φ)=0, μ (X)=1, estimates for Regularized Fuzzy.It is understood that shown in λ fuzzy mearues calculation formula such as formula (6):
μ (A ∪ B)=μ (A)+μ (B)+λ μ (A) μ (B) (6)
WhereinA ∩ B=φ, there are constant λ > -1.The value of λ can be come by solution formula (7), formula (8)
It obtains.
μi=μ ({ xi)=Crd (xi) (8)
Wherein μiIndicate notable figure xiTo the significance level of final fusion results, other measure values in power set Y can be by formula
(6) it is calculated.To ensure the efficiency of salient region detection in the present invention, it is 0 that the value of λ is taken in the present invention, it may also be said to this
Estimating used in invention, which meets classics, to be estimated.
5th step, at each pixel of notable figure, the non-negative real value in (N) fuzzy integral defined in the present invention can survey letter
Number f:X→[0,1]It is discrete-valued function, functional value is { f (x1), f (x2) ..., f (x5), wherein f (xi) be i-th (1≤i≤
5) saliency value of width notable figure.At each pixel, 5 width notable figures are ranked up so that the function f after sequence meets 0
≤f(xθ(0))≤f(xθ(1))≤…≤f(xθ(5))≤1, wherein θ are ranking functions.
6th step calculates the notable figure that (N) fuzzy integral is merged.It is understood that the definition of (N) fuzzy integral is such as public
Shown in formula (9):
Wherein meet ∨ representatives to be maximized.By can further obtain formula (10), formula in formula (9) present invention
(11):
Aθ(i)={ xθ(i), xθ(i+1)..., xθ(5), i=1,2 ..., 5 (11)
Wherein μ (xθ(i)) be and f (xθ(i)) corresponding fuzzy mearue.The value that (N) ∫ f (x) d μ are calculated is at certain
At one pixel, (N) fuzzy integral is used to merge saliency value Sal (x) after the fusion that 5 kinds of notable figures obtain, such as formula (12) institute
Show:
Sal (x)=(N) ∫ f (x) d μ (12)
7th step enhances foreground area, inhibits background area.The present invention using formula (13) come to obtained notable figure into
Row optimization obtains final notable figure.
T in the present invention1=0.1, t2=0.2, t3=0.3, t4=0.8, β1=0.25, β2=0.5.
So far the present invention has just obtained the final notable figure of 5 kinds of methods of fusion BSCA, DSR, HS, RBD, MR.
Claims (1)
1. one kind being based on the notable figure fusion method of (N) fuzzy integral, which is characterized in that steps are as follows:
The first step generates n width notable figures, n> using the n kind methods to be merged;1, it combines to obtain finite aggregate X by n width notable figures
={ x1,x2,…,xn, wherein xnIndicate the n-th width notable figure;Each width notable figure is all considered as a kind of grader, it will be notable
Property detection be considered as a kind of two classification problems, the matrix D P for obtaining decision section is indicated as shown in formula (1):
Wherein first is classified as each grader judgement pixel as the possibility of foreground, and second, which is classified as each grader judgement pixel, is
The possibility of background represents a kind of grader per a line;I-th of grader judges pixel for the probability F of foregroundiIt is expressed as Fi
=f (xi), 1≤i≤n judges pixel for the probability B of backgroundiIt is expressed as Bi=1-f (xi), wherein f (xi) it is that the i-th width is notable
The saliency value of figure, Pixel-level merge notable figure, and the saliency value of notable figure refers both to correspond to the saliency value at a certain pixel;For
This, the problem of notable figure fusion method will merge several notable figures, which is converted into, seeks corresponding (N) fuzzy integral of DP matrix first rows
The problem of;
Second step calculates related coefficient between each width notable figure, finds out similar matrix;Similarity factor dijCalculating such as formula (2) institute
Show:
Wherein , ||f(xi),f(xj)||It indicates in CIELab color spaces, the i-th width notable figure and jth width notable figure saliency value
Euclidean distance, σ2=0.1;Similarity factor dijFor describing the similarity degree between each width notable figure, dij∈[0,1], value gets over
It is big to indicate notable figure xiWith xjBetween it is more similar;
By related coefficient, obtain shown in the similar matrix such as formula (3) corresponding to n width notable figures:
Third walks, and finds out the support and confidence level between each width notable figure;Notable figure xiDegree of being supported indicate by other notable figures
Degree of support, if a width notable figure and other notable figures are all more similar, then it is assumed that their mutual support degree is also higher,
Notable figure xiDegree of being supported Sup (xi) shown in calculation formula such as formula (4):
The confidence level of notable figure reflects that certain width notable figure obtains the credibility of saliency value, and a width notable figure is by other notable figure institutes
The degree of support is higher, and the notable figure confidence level is bigger, and the confidence level of notable figure is calculated as shown in formula (5):
4th step finds out the fuzzy mearue of each width notable figure composition finite aggregate X power sets;It is the power being made of the subset of X to define Y
Collection, μ:Y→[0,1]It indicates from power set Y to [0,1]Mapping, take μ be fuzzy mearue, meet μ (φ)=0, μ (X)=1, be
Canonical;Shown in λ fuzzy mearues calculation formula such as formula (6):
μ (A ∪ B)=μ (A)+μ (B)+λ μ (A) μ (B) (6)
Wherein,There are constant λ > -1;The value of λ is obtained by solution formula (7) and formula (8);
μi=μ ({ xi)=Crd (xi) (8)
Wherein, μiIndicate notable figure xiTo the significance level of final fusion results, other measure values in power set Y by formula (6) into
Row calculates;
5th step defines the non-negative real value measurable function f in (N) fuzzy integral at each pixel of notable figure:X→[0,1]
It is discrete-valued function, functional value is { f (x1),f(x2),…,f(xn), wherein f (xn) be the n-th width notable figure saliency value;To n
Width notable figure is ranked up so that the function f after sequence meets 0≤f (xθ(0))≤f(xθ(1))≤…≤f(xθ(n))≤, 1 its
In, θ is ranking functions;
6th step calculates the notable figure that (N) fuzzy integral is merged, shown in the definition such as formula (9) of (N) fuzzy integral:
Wherein, meet ∨ representatives to be maximized, formula (10) and formula (11) are further obtained by formula (9):
Aθ(i)={ xθ(i),xθ(i+1),…,xθ(n), i=1,2 ..., n (11)
Wherein, μ (xθ(i)) be and f (xθ(i)) corresponding fuzzy mearue;The value that (N) ∫ f (x) d μ are calculated is in a certain picture
At vegetarian refreshments, (N) fuzzy integral is used to merge saliency value Sal (x) after the fusion that n kind notable figures obtain, as shown in formula (12):
Sal (x)=(N) ∫ f (x) d μ (12)
7th step enhances foreground area, inhibits background area;Obtained notable figure is optimized to obtain using formula (13)
Final notable figure
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810375979.9A CN108694710A (en) | 2018-04-18 | 2018-04-18 | One kind being based on the notable figure fusion method of (N) fuzzy integral |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810375979.9A CN108694710A (en) | 2018-04-18 | 2018-04-18 | One kind being based on the notable figure fusion method of (N) fuzzy integral |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108694710A true CN108694710A (en) | 2018-10-23 |
Family
ID=63845043
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810375979.9A Pending CN108694710A (en) | 2018-04-18 | 2018-04-18 | One kind being based on the notable figure fusion method of (N) fuzzy integral |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108694710A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111626306A (en) * | 2019-03-25 | 2020-09-04 | 北京联合大学 | Saliency map fusion method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570851A (en) * | 2016-10-27 | 2017-04-19 | 大连理工大学 | Weighted assignment D-S (Dempster-Shafer) evidence theory-based salient map fusion method |
| CN106778903A (en) * | 2017-01-09 | 2017-05-31 | 深圳市美好幸福生活安全系统有限公司 | Conspicuousness detection method based on Sugeno fuzzy integrals |
| CN106780422A (en) * | 2016-12-28 | 2017-05-31 | 深圳市美好幸福生活安全系统有限公司 | A kind of notable figure fusion method based on Choquet integrations |
-
2018
- 2018-04-18 CN CN201810375979.9A patent/CN108694710A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570851A (en) * | 2016-10-27 | 2017-04-19 | 大连理工大学 | Weighted assignment D-S (Dempster-Shafer) evidence theory-based salient map fusion method |
| CN106780422A (en) * | 2016-12-28 | 2017-05-31 | 深圳市美好幸福生活安全系统有限公司 | A kind of notable figure fusion method based on Choquet integrations |
| CN106778903A (en) * | 2017-01-09 | 2017-05-31 | 深圳市美好幸福生活安全系统有限公司 | Conspicuousness detection method based on Sugeno fuzzy integrals |
Non-Patent Citations (1)
| Title |
|---|
| SUN J等: "《Saliency Region Detection Based on Markov Absorption Probabilities》", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111626306A (en) * | 2019-03-25 | 2020-09-04 | 北京联合大学 | Saliency map fusion method and system |
| CN111626306B (en) * | 2019-03-25 | 2023-10-13 | 北京联合大学 | Saliency map fusion method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pavithra et al. | An efficient framework for image retrieval using color, texture and edge features | |
| CN108038476B (en) | A kind of facial expression recognition feature extracting method based on edge detection and SIFT | |
| Clinton et al. | Accuracy assessment measures for object-based image segmentation goodness | |
| CN104063719B (en) | Pedestrian detection method and device based on depth convolutional network | |
| CN108875600A (en) | A kind of information of vehicles detection and tracking method, apparatus and computer storage medium based on YOLO | |
| CN106570851B (en) | A kind of notable figure fusion method based on weight assignment D-S evidence theory | |
| JP2000123184A (en) | Method for detecting event in animation | |
| CN109903339B (en) | Video group figure positioning detection method based on multi-dimensional fusion features | |
| CN108960142B (en) | Pedestrian re-identification method based on global feature loss function | |
| CN109583498B (en) | Fashion compatibility prediction method based on low-rank regularization feature enhancement characterization | |
| CN106780422A (en) | A kind of notable figure fusion method based on Choquet integrations | |
| CN108830842A (en) | A kind of medical image processing method based on Corner Detection | |
| CN105893947A (en) | Bi-visual-angle face identification method based on multi-local correlation characteristic learning | |
| Li et al. | Statistical thresholding method for infrared images | |
| Zhu et al. | A modified deep neural network enables identification of foliage under complex background | |
| Liu et al. | Automatically face detection based on BP neural network and Bayesian decision | |
| Tao et al. | Local difference ternary sequences descriptor based on unsupervised min redundancy mutual information feature selection | |
| CN108320281A (en) | A kind of image significance detection method and terminal based on multiple features diffusion | |
| Al-Mohair et al. | Color space selection for human skin detection using color-texture features and neural networks | |
| Wang et al. | SpecVAT: Enhanced visual cluster analysis | |
| CN105550642B (en) | Gender identification method and system based on multiple dimensioned linear Differential Characteristics low-rank representation | |
| CN106886771A (en) | The main information extracting method of image and face identification method based on modularization PCA | |
| CN110414336A (en) | A kind of depth complementation classifier pedestrian's searching method of triple edge center loss | |
| CN108694710A (en) | One kind being based on the notable figure fusion method of (N) fuzzy integral | |
| CN112446340A (en) | Pedestrian searching method and system combining pedestrian local feature and decorative attribute classification and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181023 |
|
| RJ01 | Rejection of invention patent application after publication |