CN102103744A - Image segmentation method based on minimum spanning trees and statistical learning theory - Google Patents
Image segmentation method based on minimum spanning trees and statistical learning theory Download PDFInfo
- Publication number
- CN102103744A CN102103744A CN 201110031892 CN201110031892A CN102103744A CN 102103744 A CN102103744 A CN 102103744A CN 201110031892 CN201110031892 CN 201110031892 CN 201110031892 A CN201110031892 A CN 201110031892A CN 102103744 A CN102103744 A CN 102103744A
- Authority
- CN
- China
- Prior art keywords
- image
- limit
- segmentation
- minimum spanning
- statistical learning
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an image segmentation method based on minimum spanning trees and a statistical learning theory. On the basis of image map model construction, a bottom-up combination strategy is adopted, and the process of generating a plurality of minimum spanning trees is finished according to statistical learning theory-based synthesis criterions designed by the invention by using a minimum spanning tree algorithm. The segmentation method can effectively utilize regional statistical characteristics at the same time of meeting global optimum, is suitable for the high-resolution segmentation of various images, and achieves relatively higher noise immunity and relatively better segmentation effects on texture regions; and in the method, good regional boundaries can be obtained simultaneously.
Description
Technical field
This method belongs to the image processing and pattern recognition field, particularly relates to a kind of new OO image partition method based on minimum spanning tree Optimum Theory and Statistical Learning Theory.
Background technology
High spatial resolution remote sense image provides the high precision space geometry information of ground landscape, abundant texture information and multispectral information for us, make that traditional classification of remote-sensing images method based on pixel is inapplicable, thereby make high-resolution remote sensing image handle to face the challenge of the details that image provides.For this reason, Baatz and Schape pointed out in [1] in 1999: important semantic interpretation more needs to represent with the mutual relationship between object and the object in the significant image rather than with pixel one by one, therefore, OO high-resolution remote sensing image Target Recognition and sorting technique have been proposed, promptly at first image is cut apart the formation object zone, and come description object with hierarchical network, be that unit carries out Target Recognition again with the object.Blaschke and Strobl send " what is wrong for pixel? " in [1] query, point out that traditional multidimensional feature space classification based on pixel does not utilize spatial information, the high resolution image that particularly may belong to same soil cover type for neighbor, to be divided into the classification on basis, be better than the method for each pixel of tradition under many circumstances.The new sorter proof of object-based spectrum, shape, texture, spatial relationship and people's the further reasoning of knowledge is very useful in the high spatial resolution field, and it meets the rule of human recognition objective, has improved nicety of grading and details.Developing rapidly of object-based research field makes the geography information scientific domain produce an object-based image analysing computer (Object-Based Image Analysis again, OBIA) sub-subject, it is specialized in and automatically remote sensing image is divided into significant object, assess their feature by space, spectrum and time scale, to produce the new geography information [2] of GIS data layout.OO image segmentation is one of main method of obtaining object, OO target identification technology is by the spectrum to object, how much, texture, the incompatible recognition objectives of message block such as spatial neighborhood relation, from theoretical and practical application, find, the quality of Object Segmentation directly influences the effect and the precision of image classification identification, can be accurate because it is directly connected to, the geological information and the structural information of target in extracting on the image effectively, therefore, the object-oriented high-resolution remote sensing image is partitioned into crucial and basic in the remote sensing image processing, and the research of object-oriented dividing method also becomes one of high spatial resolution focus and difficult point.
Image segmentation has been a classical problem of Flame Image Process, computer vision, cuts apart from the multiple goal that is divided into of early stage single goal and background; Be divided into object level from Pixel-level and cut apart, be divided into definite target from uncertain target and cut apart, the researchist has proposed a large amount of partitioning algorithms according to demand.The elementary tactics of image segmentation is based on two fundamental characteristics of grey scale pixel value: uncontinuity and similarity.The pixel of intra-zone generally has certain similarity, and generally has certain uncontinuity on the border between the zone.So can being divided in view of the above, partitioning algorithm utilize interregional characteristic discontinuous based on the algorithm on border and the algorithm based on the zone of the regional interior similarity of utilization.
Development along with image Segmentation Technology, people have turned to from initial simple Threshold Segmentation, rim detection and extracted region Study of Image Segmentation new notion, new method have been introduced the image segmentation field, pay much attention to effective combination of multiple partitioning algorithm.By two or more method of comprehensive use, can partly overcome the problem that independent image segmentation algorithm is difficult to general pattern is obtained gratifying segmentation effect, and take which type of combination could embody the advantage of the whole bag of tricks, remedy deficiency separately, the Expected Results of obtaining is still one of subject matter of people's concern.After the 1980s, more and more scholars begins achievements in research such as fuzzy theory, Markov model, Genetic Algorithms Theory, fractal theory and wavelet theory are applied to Study of Image Segmentation, and has obtained remarkable progress.But because the complicated diversity of image type though people have carried out research extensively and profoundly, does not still have the general image segmentation theory at present and proposes, existing algorithm all respectively has limitation.Therefore, exploring new segmentation theory and partitioning algorithm is significant for Flame Image Process, analysis
Http:// www.tu-dresden.de/ioer/statisch/segmentation-evaluation/ index.html is one and is specifically designed to the website that the remote sensing image segmentation software is estimated, simply introduced various OO high-resolution remote sensing image segmentation software methods and performance in the website, showed the segmentation result of each software high resolving power city and suburb image.That shows at present mainly contains following software: BerkeleyImgSeg, Definens Developer, CAEASR, Data Dissection Tool, Edge Detection and Image SegmentationON (EDISON) System, ENVI Feature Extraction, Extended Watershed EWS (Multi-channel watershed transformation), InfoPack, MinimumEntropy Approach, Image segmentation for Erdas Imagine, Imagine WS for Erdas Imagine, PABAT 0.32, RHSEG, SCRM, SegSAR, HaIcon Seg, the dividing method that is adopted of every kind of software is described below:
Definens Developer and BerkeleyImgSeg are that two object-oriented that the remote sensing field is released is at present cut apart and classification software, the thought of these two segmentation software is identical, it is cut apart thought and all derives from document [3], all adopt region growing and consolidation strategy, the weight and the region shape characteristic of multiband light spectral property have been considered in its merging criterion, realize multiple dimensioned, multi-level image is cut apart, the upper strata is to liking its merging of one deck object down, image is carried out hierarchical description, every layer of segmentation result all added up its spectrum, texture, shape, the topological characteristic of object, the attribute information that concerns between the levels object, this statistics can be used for further image classification and analysis.
Two softwares of CAEASR and InfoPack are according to document [4], suppose to have in the image r and obey the zone that Gamma distributes, and by the definition cost function, (Simulated Annealing, SA) method is finished image segmentation to adopt simulated annealing.Simulated annealing method is S.Kirkpatrick, a kind of general probabilistic algorithm that C.D.Gelatt and M.P.Vecchi invented in nineteen eighty-three, the optimum solution that is used for looking for proposition in a big search space is a kind of general optimized Algorithm, optimization problems such as optimum control, machine learning, neural network have been widely used at present, algorithm is done initial with arbitrfary point in the search space earlier, each step is selected one " neighbours " earlier, and then calculates the probability that arrives " neighbours " from existing position.
Data Dissection Tool adopts the superparamagnetism cluster algorithm that proposes in [5], and (Super ParamagneticClusting SPC) realizes image segmentation and analysis.Super paramagnetic clustering method is a kind of brand-new thought, it is with the physical thought message area that induces one, the thermodynamics of non-homogeneous Potts model is assembled motion regard data clusters as, promptly in certain temperature range, data are in super paramagnetic phase place, utilize the correlativity between the data point to come cluster then.This algorithm is not strong to the dependence of various parameters, is mainly used in cluster, image analysing computer, statistical physics.
Edge Detection and Image SegmentationON (EDISON) System, be based on the Mean shift algorithm that proposes in the document [6] and finish cutting apart chromatic image, the estimation that this notion of Mean Shift is proposed in [7] by people such as Fukunaga the earliest about the probability density gradient function, it is a kind of nonparametric feature spatial analysis technology, this algorithm is the process of an iteration, promptly calculate the skew average of current point earlier, move this and put its skew average, then as new starting point, continue to move, up to the end that meets some requirements.Mean Shift algorithm is the method for an adaptive gradient rising search peak in essence, and Mean Shift mainly is used in detection and three aspects of optimization of cluster, mode, has been widely used in the cluster of computer vision and Flame Image Process.
ENVI Feature Extraction introduces lambda parameter control on document [8] and Mumford-Shah model basis cuts apart yardstick and realizes quick multi-scale division [9] to image by the region growing mode.
The hyperchannel that EWS proposes based on document [10] is expanded the dividing method of watershed transform.Minimum Entropy Approach adopts triangulation (Triangulation), Image segmentation for Erdas Imagine, Imagine WS forErdas Imagine, PABAT32 all to adopt the region growing method, and PABAT32 has considered the remote sensing image uncertainty according to document [11] in cutting apart; RHSEG adopts level region growing (Hierarchical region growing, [12]); SCRM adopts watershed divide combine with region growing (Watershed and region growing, [13]), SegSAR, HaIcon Seg jointing edge and regional cut apart (Hybrid edge/region oriented).
Sum up these methods, can find that OO image partitioning algorithm mainly concentrates on region growing and watershed divide, introduce the multi-scale division of simulated annealing, Mean Shift, superparamagnetism cluster algorithm, Mumford-Shah, comprehensive zone and local edge simultaneously.This has illustrated cuts apart algorithm in region growing, division and merging, hierarchical clustering and watershed divide etc. to combine other criterions is main modes that the OO high-resolution remote sensing image multi-scale division of current realization adopts, and the research emphasis of most of document is new theory to be joined cut apart the criterion setting and realize that multiple dimensioned image cuts apart with optimizing.
In partitioning algorithm based on region growing, the selection of seed points and the setting of merging criterion are the committed steps that influences segmentation effect, the center that seed points should be selected in homogeneous region is best, and to select and obtain the homogeneity center be very difficult as seed points, and the characteristics such as multispectral, shape, texture that how to make full use of image also are that this method need be considered in cutting apart criterion.The watershed segmentation algorithm also is a kind of OO image segmentation algorithm of widespread use, but its weak point is to occur the over-segmentation phenomenon easily.Therefore, a lot of scholar's research also propose solution to this problem, as gradient image being carried out pre-service reducing noise, overdivided region is merged and the gradient of input picture is got methods such as threshold value.
The background technology citing document:
1.Blaschke,T.andJ.Strobl,What′s?wrong?with?pixels?Some?recent?developments?interfacing?remote?sensing?andGIS.GIS-Zeitschrift?für?Geoinformationssysteme,2001.14(6):p.12-17.
2.Blaschke,T.,S.Lang,and?G.J.Hay,eds.Object-Based?Image?Analysis:Spatial?Concepts?forKnowledge-Driven?Remote?Sensing?Applications.Lecture?Notes?in?Geoinformation?and?Cartography,ed.W.Cartwright,G.Gartner,L.Meng,et?al.2008,Springer-Verlag?Berlin?Heide1berg.
3.Baatz,M.andA.
Multiresolution?Segmentation:an?optimization?approach?for?high?quality?multi-scaleimage?segmentation?in?Strobl,J.,Blaschke,T.,Griesebner,G.(eds),Angewandte?GeographischeInformationsverarbeitung?XII.2000.Wichmann,Heidelberg.
4.Cook,R.,I.McConnell,D.Stewart,and?C.J.Oliver.Segmentation?and?simulated?annealing.in?MicrowaveSensing?and?Synthetic?Aperture?Radar?1996.Taormina,Italy?SPIE.
5.Ferbe?r,C.v.and?F.
Cluster?update?algorithm?and?recognition.Physical?Review?E,2000.62(2):p.Rl461-R1464.
6.Comaniciu,D.and?P.Meer,Mean?shift:a?robust?approach?toward?feature?space?analysis.Pattern?Analysis?andMachine?Intelligence,IEEE?Transactions?on,2002.24(5):p.603-619.
7.Fukunaga,K.and?L.D.Hostetle?r,The?estimation?of?the?gradient?of?a?density?function,with?applications?inpattern?recognition.Information?Theory,IEEE?Transactions?on,1975.21(1):p.32-40.
8.Koepfler,G.,C.Lopez,and?J.M.Morel,A?multiscale?algorithm?for?image?segmentation?by?variational?method.SIAM?Journal?on?Numerical?Analysis?1994.31(1):p.282-299.
9.Robinson,D.J.,N.J.Redding,and?D.J.Crisp.Implementation?of?a?fast?algorithm?for?segmenting?SAR?imagery.[Technical?Report(Defence?Science?and?Technology?Organisation(Australia))]2002-01.
10.Li,P.and?X.Xiao,Multispectral?image?segmentation?by?a?multichannel?watershed-based?approach.International?Journal?of?Remote?Sensing,2007.28(19):p.4429-4452.
11.Lucieer,A.,Uncertainties?in?Segmentation?and?their?Visualisation.October?2004,Utrecht?University?andInternational?Institute?for?Geo-Information?Science?and?Earth?Observation(ITC).
12.Tilton,J.C.Analysis?of?hierarchically?related?image?segmentations.in?Proc?IEEE?Workshop?on?Advances?inTechniques?for?Analysis?of?Remotely?Sensed?Data.2003.
13.Castilla,G.,G.J.Hay,and?J.R.Ruiz-Gallardo,Size-constrained?region?merging(SCRM):an?automateddelineation?tool?for?assisted?photo?interpretation.Photogrammetric?Engineering?and?Remote?Sensing,2008.74(4):p.409-419.
Summary of the invention
At above-mentioned difficulties and problem, the present invention adopt a kind of based on the minimum spanning tree theory image segmentation and the criterion of cutting apart based on the Statistical Learning Theory design is proposed, realized OO image segmentation, avoid the seed points in the region growing to select problem, and efficiently solved the over-segmentation problem.
Image partition method based on minimum spanning tree and Statistical Learning Theory provided by the present invention may further comprise the steps:
Step 1, image is represented with the simple graph model that a pixel of each summit correspondence image of simple graph model connects with the limit between per two adjacent vertexes, the limit weights on every limit are the difference between corresponding two pixels in two summits of this limit connection;
Step 2, based on Statistical Learning Theory setting regions merging criterion, and partitioning parameters is set;
Step 3, step 1 gained simple graph model is carried out cutting apart based on the image of minimum spanning tree, minimum weights limit when cutting apart from the simple graph model begins to merge up to maximum weights limit by regional merging criterion, merges the zone that each minimum spanning tree that generates is represented a connection.
And weights account form in limit is in the step 1, two summits the difference between corresponding two pixels adopt wave band weighted quadratic distance and calculate.
And, be that as a learning process, learning process is to regard image segmentation as a regression estimation problem with the image segmentation process described in the step 2 based on Statistical Learning Theory setting regions merging criterion specific implementation; According in the Statistical Learning Theory based on the consistent learning algorithm setting regions merging criterion of stablizing of the minimized β of empiric risk, the empiric risk minimization problem is converted into a probability problem, and the loss function in the statistical learning is corresponding with limit weights account form in the step 1.
And the partitioning parameters of step 2 comprises cuts apart scale parameter, and the value effect step 3 image of cutting apart scale parameter by setting is cut apart the size in gained zone.Cut apart scale parameter according to maximum limit weights setting that the zone allowed.
And the partitioning parameters of step 2 comprises the Minimum Area area parameters, prevents that by the numerical value that the Minimum Area area parameters is set the step 3 image from generating the zonule when cutting apart.
And the described image segmentation based on minimum spanning tree of step 3 adopts the Kruskal minimal spanning tree algorithm to realize.
The present invention represents image with graph model, seek homogeneity district problem in the image segmentation and be converted into minimum spanning tree problem, simultaneously the image segmentation process is regarded as a learning process; Adopt the Kruskal minimal spanning tree algorithm, and in conjunction with its algorithm characteristic, the limit weight function is corresponding with loss function, Statistical Learning Theory is melted into the minimum spanning tree image segmentation.This dividing method can effectively utilize the range statistics characteristic when satisfying global optimum, be suitable for all kinds of high resolution images and cut apart; Have noise robustness preferably, also can obtain segmentation effect preferably, can obtain good zone boundary simultaneously texture region.
Description of drawings
Fig. 1 is to the synoptic diagram of simple graph model representation based on top-down minimum spanning tree segmentation result, wherein Fig. 1 a is the gray-scale value of a 5*5 image, Fig. 1 b represents that for the graph model corresponding with Fig. 1 a Fig. 1 c is the minimum spanning tree of Fig. 1 b, and Fig. 1 d is the segmentation result to Fig. 1 c.
Fig. 2 is the synoptic diagram of threshold affects segmentation result, and wherein Fig. 2 a is the segmentation result that threshold value obtains when getting higher value, and Fig. 2 b is the segmentation result that threshold value obtains when getting smaller value.
Fig. 3 is the changes of threshold rule synoptic diagram of the embodiment of the invention.
Embodiment
In conjunction with the accompanying drawings and embodiments of the invention, technical solution of the present invention is elaborated.The implementation procedure of embodiment is as follows:
Step 1, image is represented with the simple graph model that a pixel of each summit correspondence image of simple graph model connects with the limit between per two adjacent vertexes, the limit weights on every limit are the difference between corresponding two pixels in two summits of this limit connection.
The simple graph model is a kind of in the existing pyramid model.The limit weights are the weight on every limit in the simple graph model, and embodiment adopts wave band weighted quadratic distance and calculates weight.Concrete account form is a prior art, and the present invention will not give unnecessary details.
Step 2, based on Statistical Learning Theory setting regions merging criterion, and partitioning parameters is set.
Embodiment is that as a learning process, learning process is to regard image segmentation as a regression estimation problem with the image segmentation process based on Statistical Learning Theory setting regions merging criterion specific implementation; According in the Statistical Learning Theory based on the consistent learning algorithm setting regions merging criterion of stablizing of the minimized β of empiric risk, the empiric risk minimization problem is converted into a probability problem, and the loss function in the statistical learning is corresponding with limit weights account form in the step 1.
The b uniform stability algorithm of an element of change that proposes according to Bousquet and E1isseeff and the extensive error bound of deriving according to the McDiarmid inequality thereof, according to Statistical Learning Theory thought, through deriving, designed following criterion:
Wherein, S
1And S
2Represent current two zones, n
1And n
2The expression region S
1And S
2In pixel count (number of vertex), n represents to work as front, w
nBe to work as the front in the merging process from small to large by the limit weights, it connects S
1And S
2Two zone , ﹠amp of expression; ﹠amp; Expression is satisfied simultaneously.If P is (S
1, S
2) then merge region S for TRUE
1And S
2, if P (S
1, S
2) be then nonjoinder of FALSE.
This criterion is by three parameter (n
i, δ M) controls threshold value, and just in time is the parameter that yardstick is cut apart in control.As shown in Figure 3, get δ=0.1 (being labeled as delta=0.1 among the figure), horizontal ordinate is n
i(area size), ordinate are that th (pass through by this threshold value
Calculate).Along y direction, be ascending 30,40,50,80,100,120 o'clock the change curve of different value of getting of M; Along X direction, the changes of threshold trend when being the area increase.n
iBe the pixel count of regional i, it is worth with the merging process dynamic change; Probability δ is a little value between 0 and 1, and its variation is little to threshold affects, has the fine setting effect, and according to the analysis of probability theory and front, we establish its value usually is 0.1; M is the loss upper bound, and it is corresponding to the supremum of limit power, and we can select its value according to maximum limit weights (maximum difference in the zone).
In the present embodiment, δ fixedly is taken as 0.1, n
iAutomatically change with merging process, therefore, only get M and cut apart the parameter of yardstick,, can obtain the segmentation result of image under different scale by the different scale parameters of cutting apart is set as control.The threshold value that big M value obtains is bigger, and the extensive error of permission is big, promptly allows intra-zone to change greatly, thereby obtains the segmentation result of large scale.
During concrete enforcement, also can adopt other regional merging criterions based on Statistical Learning Theory, for example based on Statistical Learning Theory by the criterion of the design of all pixel characteristic of zone etc.
During partitioning parameters is provided with, generated the zonule when cutting apart, the present invention proposes to comprise in the partitioning parameters parameter that is used to control Minimum Area area size.When certain region area when cutting apart, occurring less than Minimum Area area size parameter, that can carry out is treated to: begin in proper order from small to large by the limit weights again, if when two zones that the front connected do not belong to same zone, and certain region area in two zones that this limit connected then merges these two zones less than specifying the Minimum Area area.Begin to increase up to maximum weights limit by the minimum weights limit from the simple graph model again, guaranteed similarity maximum between the merging zone like this by regional merging criterion.
Step 3, step 1 gained simple graph model is carried out cutting apart based on the image of minimum spanning tree, minimum weights limit when cutting apart from the simple graph model begins to merge up to maximum weights limit by regional merging criterion, merges the zone that each minimum spanning tree that generates is represented a connection.
Minimum spanning tree problem is described as seeking limit power and minimum generation tree, and for image segmentation, the connected region problem of seeking the difference minimum just can be converted into minimum spanning tree problem.Therefore, after relation is used the simple graph model representation between step 1 has realized the pixel of graphical representation, can carry out image to step 1 gained simple graph model based on minimum spanning tree cuts apart, referring to Fig. 1, wherein Fig. 1 a is the gray-scale value of a 5*5 image, Fig. 1 b represents that for the graph model corresponding with Fig. 1 a Fig. 1 c is the minimum spanning tree of Fig. 1 b, and Fig. 1 d is the segmentation result to Fig. 1 c.The scale parameter of cutting apart in the step 2 setting influences segmentation result, and referring to Fig. 2, Fig. 2 a is the segmentation result that threshold value obtains when getting higher value, and Fig. 2 b is the segmentation result that threshold value obtains when getting smaller value.Mainly contain top-down division minimum spanning tree and bottom-up merging generates two kinds of strategies of a plurality of minimum spanning trees based on the image segmentation of minimum spanning tree.Embodiment adopts Kruska1 minimal spanning tree algorithm and bottom-up consolidation strategy to realize cutting apart based on the image of minimum spanning tree, and specific algorithm and strategy are prior art, and the present invention will not give unnecessary details.
Claims (7)
1. based on the image partition method of minimum spanning tree and Statistical Learning Theory, be characterized in: may further comprise the steps,
Step 1, image is represented with the simple graph model that a pixel of each summit correspondence image of simple graph model connects with the limit between per two adjacent vertexes, the limit weights on every limit are the difference between corresponding two pixels in two summits of this limit connection;
Step 2, based on Statistical Learning Theory setting regions merging criterion, and partitioning parameters is set;
Step 3, step 1 gained simple graph model is carried out cutting apart based on the image of minimum spanning tree, minimum weights limit when cutting apart from the simple graph model begins to merge up to maximum weights limit by regional merging criterion, merges the zone that each minimum spanning tree that generates is represented a connection.
2. image partition method according to claim 1 is characterized in that: weights account form in limit is in the step 1, two summits the difference between corresponding two pixels adopt wave band weighted quadratic distance and calculate.
3. image partition method according to claim 2, it is characterized in that: described in the step 2 be based on Statistical Learning Theory setting regions merging criterion specific implementation, as a learning process, learning process is to regard image segmentation as a regression estimation problem with the image segmentation process; According in the Statistical Learning Theory based on the minimized consistent learning algorithm setting regions merging criterion of stablizing of empiric risk, the empiric risk minimization problem is converted into a probability problem, and the loss function in the statistical learning is corresponding with limit weights account form in the step 1.
4. image partition method according to claim 1 is characterized in that: the partitioning parameters of step 2 comprises cuts apart scale parameter, and the value effect step 3 image of cutting apart scale parameter by setting is cut apart the size in gained zone.
5. image partition method according to claim 4 is characterized in that: cut apart scale parameter according to maximum limit weights setting that the zone allowed.
6. image partition method according to claim 1 is characterized in that: the partitioning parameters of step 2 comprises the Minimum Area area parameters, prevents that by the numerical value that the Minimum Area area parameters is set the step 3 image from generating the zonule when cutting apart.
7. image partition method according to claim 1 is characterized in that: the described image segmentation based on minimum spanning tree of step 3, adopt the Kruskal minimal spanning tree algorithm to realize.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110031892 CN102103744A (en) | 2011-01-28 | 2011-01-28 | Image segmentation method based on minimum spanning trees and statistical learning theory |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110031892 CN102103744A (en) | 2011-01-28 | 2011-01-28 | Image segmentation method based on minimum spanning trees and statistical learning theory |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102103744A true CN102103744A (en) | 2011-06-22 |
Family
ID=44156488
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110031892 Pending CN102103744A (en) | 2011-01-28 | 2011-01-28 | Image segmentation method based on minimum spanning trees and statistical learning theory |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102103744A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289797A (en) * | 2011-08-02 | 2011-12-21 | 于元隆 | Method for hierarchically segmenting image based on Bhattacharyya distance |
CN103733207A (en) * | 2011-08-15 | 2014-04-16 | 德米特里·瓦莱里维奇·施穆克 | Method of image segmentation |
CN112927257A (en) * | 2021-03-22 | 2021-06-08 | 哈尔滨理工大学 | Stamping part defect image segmentation method based on improved MRF |
CN114931388A (en) * | 2022-04-26 | 2022-08-23 | 广东医科大学 | Neuron spike potential classification method and device based on parallel superparamagnetic clustering algorithm, storage medium and computer equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006076312A2 (en) * | 2005-01-10 | 2006-07-20 | Cytyc Corporation | Method for improved image segmentation |
CN101065774A (en) * | 2004-11-26 | 2007-10-31 | 史诺伟思有限公司 | Image segmentation |
CN101826208A (en) * | 2010-04-26 | 2010-09-08 | 哈尔滨理工大学 | Image segmentation method combining support vector machine and region growing |
-
2011
- 2011-01-28 CN CN 201110031892 patent/CN102103744A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101065774A (en) * | 2004-11-26 | 2007-10-31 | 史诺伟思有限公司 | Image segmentation |
WO2006076312A2 (en) * | 2005-01-10 | 2006-07-20 | Cytyc Corporation | Method for improved image segmentation |
CN101826208A (en) * | 2010-04-26 | 2010-09-08 | 哈尔滨理工大学 | Image segmentation method combining support vector machine and region growing |
Non-Patent Citations (2)
Title |
---|
《中国优秀硕士学位论文全文数据库信息科技辑》 20101015 李涟凤 基于支持向量机的医学图像分割 第3页第2行至第4行,第6页第24行至29行,第11页以及摘要 1-7 , 2 * |
《北京交通大学学报》 20070831 谭衢霖, 刘正军, 沈伟 一种面向对象的遥感影像多尺度分割方法 全文 1-7 第31卷, 第4期 2 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289797A (en) * | 2011-08-02 | 2011-12-21 | 于元隆 | Method for hierarchically segmenting image based on Bhattacharyya distance |
CN102289797B (en) * | 2011-08-02 | 2013-01-23 | 于元隆 | Method for hierarchically segmenting image based on Bhattacharyya distance |
CN103733207A (en) * | 2011-08-15 | 2014-04-16 | 德米特里·瓦莱里维奇·施穆克 | Method of image segmentation |
CN112927257A (en) * | 2021-03-22 | 2021-06-08 | 哈尔滨理工大学 | Stamping part defect image segmentation method based on improved MRF |
CN114931388A (en) * | 2022-04-26 | 2022-08-23 | 广东医科大学 | Neuron spike potential classification method and device based on parallel superparamagnetic clustering algorithm, storage medium and computer equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Benz et al. | Multi-resolution, object-oriented fuzzy analysis of remote sensing data for GIS-ready information | |
Zhang et al. | Heuristic sample learning for complex urban scenes: Application to urban functional-zone mapping with VHR images and POI data | |
Zhao et al. | Automatic recognition of loess landforms using Random Forest method | |
CN102096816B (en) | Multi-scale multi-level image segmentation method based on minimum spanning tree | |
Dong et al. | Selection of LiDAR geometric features with adaptive neighborhood size for urban land cover classification | |
CN111611960B (en) | Large-area ground surface coverage classification method based on multilayer perceptive neural network | |
Na et al. | Object‐based large‐scale terrain classification combined with segmentation optimization and terrain features: A case study in China | |
Buyukdemircioglu et al. | Deep learning based roof type classification using very high resolution aerial imagery | |
Ling et al. | Learning-based superresolution land cover mapping | |
CN102103744A (en) | Image segmentation method based on minimum spanning trees and statistical learning theory | |
Zhou et al. | Stratified Object‐Oriented Image Classification Based on Remote Sensing Image Scene Division | |
Cremon et al. | Geological and terrain attributes for predicting soil classes using pixel-and geographic object-based image analysis in the Brazilian Cerrado | |
Sutha | Object based classification of high resolution remote sensing image using HRSVM-CNN classifier | |
Matci et al. | Optimization-based automated unsupervised classification method: A novel approach | |
Wang et al. | Local scale-guided hierarchical region merging and further over-and under-segmentation processing for hybrid remote sensing image segmentation | |
Qian et al. | Spatial contextual noise removal for post classification smoothing of remotely sensed images | |
Lin et al. | Slope unit-based genetic landform mapping on Tibetan plateau-a terrain unit-based framework for large spatial scale landform classification | |
Straumann et al. | Delineation of valleys and valley floors | |
CN107392926B (en) | Remote sensing image feature selection method based on early-stage land thematic map | |
Ghamisi et al. | Extending the fractional order Darwinian particle swarm optimization to segmentation of hyperspectral images | |
Nikfar et al. | Optimization of multiresolution segmentation by using a genetic algorithm | |
Nurwauziyah et al. | Satellite image classification using decision tree, SVM and k-nearest neighbor | |
Zhao et al. | MarsMapNet: A Novel Superpixel-Guided Multi-view Feature Fusion Network for Efficient Martian Landform Mapping | |
Jiang et al. | Object-oriented building extraction by DSM and very highresolution orthoimages | |
Cui et al. | An effective graph-based hierarchy image segmentation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20110622 |