CN103514599B - A kind of segmentation of the image optimum based on neighborhood total variation scale selection method - Google Patents
A kind of segmentation of the image optimum based on neighborhood total variation scale selection method Download PDFInfo
- Publication number
- CN103514599B CN103514599B CN201310386335.7A CN201310386335A CN103514599B CN 103514599 B CN103514599 B CN 103514599B CN 201310386335 A CN201310386335 A CN 201310386335A CN 103514599 B CN103514599 B CN 103514599B
- Authority
- CN
- China
- Prior art keywords
- image
- index
- segmentation
- result images
- segmentation result
- 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.)
- Active
Links
Abstract
The invention discloses a kind of image optimum based on neighborhood total variation segmentation scale selection method.Comprise the steps: 1) image is split with different segmentation yardsticks, obtain a series of Image Segmentation result images; 2) according to the homogeney index of object inside in Image Segmentation result images under homogeney evaluation method calculating different scale of the present invention; 3) according to the heterogeneity index under heterogeneous evaluation method calculating different scale according to the present invention between Image Segmentation result images object; 4) by the normalization of homogeney exponential sum heterogeneity index, specify the weight of heterogeneity index, segmentation yardstick corresponding to comprehensive evaluation index minimum value is the best segmental scale of image.The present invention's application neighborhood total variation is evaluated Image Segmentation result, consider the heterogeneity between the homogeney of Image Segmentation object inside and object, for image optimum segmentation yardstick provides objective system of selection, be conducive to provide the segmentation effect of image and the subsequent treatment of image.
Description
Technical field
The present invention relates to the technical field of the best segmental scale how selecting remote sensing image, particularly relate to a kind of image optimum based on neighborhood total variation segmentation scale selection method.
Background technology
Image Segmentation is the committed step of object-oriented image processing, and the quality of segmentation directly can affect the subsequent treatment effect of image, and the selection wherein splitting yardstick has very important effect in the segmentation effect of image.Best segmental scale is defined as after image utilizes this multi-scale segmentation, and ground class can be expressed by one or several object, object size and ground object target size close, the gray scale of imaged object inside is relatively more even, and the gray difference between object is larger.Selection at present about the best segmental scale of image has multiple method, as object and neighborhood absolute mean difference variance ratio (RatioofMeanDifferencetoNeighborstoStandardDeviation, RMAS), vector distance method, but these two kinds of methods are biased to some extent, can not consider the problem of image less divided or over-segmentation more all sidedly.The segmentation effect that Espindola analyzes image by spatial auto-correlation carries out image optimum segmentation scale selection, but choice criteria wherein can not embody the optimality of Image Segmentation well.
Summary of the invention
The object of the invention is the difficult problem solving Image Segmentation scale selection, provide a kind of method based on neighborhood total variation to carry out best segmental scale selection.
Image optimum segmentation scale selection method based on neighborhood total variation comprises the steps:
1) image is split with different segmentation yardsticks, obtain a series of Image Segmentation result images;
2) the homogeney index H of object inside in the Image Segmentation result images under each segmentation yardstick is calculated according to homogeney evaluation method;
3) the heterogeneity index I between object in the Image Segmentation result images under each segmentation yardstick is calculated according to heterogeneous evaluation method;
4) by homogeney index H and heterogeneity index I normalization, specify the weight of heterogeneity index I, calculate the comprehensive evaluation index F (H of the Image Segmentation result images under each yardstick, I), obtain Image Segmentation comprehensive evaluation index F (H, I) minimum value, the segmentation yardstick that the minimum value of comprehensive evaluation index F (H, I) is corresponding is the best segmental scale of image.
Described step 2) comprising:
(1) total number n of cutting object in each Image Segmentation result images is obtained;
(2) the area a of a kth cutting object in each Image Segmentation result images is set
kwith gray standard deviation v
k;
(3) the homogeney index H of all cutting object inside in each Image Segmentation result images is tried to achieve by following formula,
Described step 3) comprises:
(1) mould of the gradient g of raw video f (x, y) is tried to achieve by following formula | g (f (x, y)) |,
(2) setting cutting object bound radius contiguous range radius in each Image Segmentation result images is r, and contiguous range is
, obtain contiguous range
on number of pixels
(3) contiguous range in each Image Segmentation result images is tried to achieve by following formula
the mould of upper image gradient | g (f (x, y)) | integration
(4) the heterogeneity index I in each Image Segmentation result images between imaged object is tried to achieve by following formula,
Described step 4) comprises:
(1) the homogeney index H of all Image Segmentation result images and heterogeneity index I is distinguished normalization by two formulas of pressing,
(2) specify the weight ρ of heterogeneity index I, by homogeney index H and the heterogeneity index I of following formula combined imaging segmentation result image, try to achieve the Image Segmentation evaluation number F (H, I) of each Image Segmentation result images,
F(H,I)=(1-ρ)F(H)+ρF(I);
(3) ask the minimum value of Image Segmentation comprehensive evaluation index F (H, I), segmentation yardstick corresponding to the minimum value of comprehensive evaluation index F (H, I) is the best segmental scale of image.
The beneficial effect that the present invention compared with prior art has:
(1) heterogeneity between the homogeney of Image Segmentation object inside and object has been considered, comprehensively objective to the evaluation analysis of Image Segmentation quality;
(2) homogeney index calculation method of the present invention is subject to the restriction of standard deviation, area and object sum simultaneously, can consider the homogeney of image from image on the whole, the over-segmentation in energy objective measure Image Segmentation and front segmentation phenomenon;
(3) heterogeneity index computing application neighborhood total variation of the present invention, significantly can embody the optimality of heterogeneity between object and Image Segmentation yardstick, for Image Segmentation provides best segmental scale choice criteria.
Accompanying drawing illustrates:
Fig. 1 is the segmentation result figure carrying out splitting under 10 different scales of the present invention;
Fig. 2 is the curve map of Image Segmentation yardstick evaluation number of the present invention and segmentation yardstick.
Embodiment:
Image optimum segmentation scale selection method based on neighborhood total variation comprises the steps:
1) image is split with different segmentation yardsticks, obtain a series of Image Segmentation result images;
2) the homogeney index H of object inside in the Image Segmentation result images under each segmentation yardstick is calculated according to homogeney evaluation method;
3) the heterogeneity index I between object in the Image Segmentation result images under each segmentation yardstick is calculated according to heterogeneous evaluation method;
4) by homogeney index H and heterogeneity index I normalization, specify the weight of heterogeneity index I, calculate the comprehensive evaluation index F (H of the Image Segmentation result images under each yardstick, I), obtain Image Segmentation comprehensive evaluation index F (H, I) minimum value, the segmentation yardstick that the minimum value of comprehensive evaluation index F (H, I) is corresponding is the best segmental scale of image.
Described step 2) comprising:
(1) total number n of cutting object in each Image Segmentation result images is obtained;
(2) the area a of a kth cutting object in each Image Segmentation result images is set
kwith gray standard deviation v
k;
(3) the homogeney index H of all cutting object inside in each Image Segmentation result images is tried to achieve by following formula,
Described step 3) comprises:
(1) mould of the gradient g of raw video f (x, y) is tried to achieve by following formula | g (f (x, y)) |,
(2) setting cutting object bound radius contiguous range radius in each Image Segmentation result images is r, and contiguous range is
obtain contiguous range
on number of pixels
(3) contiguous range in each Image Segmentation result images is tried to achieve by following formula
the mould of upper image gradient | g (f (x, y)) | integration,
(4) the heterogeneity index I in each Image Segmentation result images between imaged object is tried to achieve by following formula,
Described step 4) comprises:
(1) the homogeney index H of all Image Segmentation result images and heterogeneity index I is distinguished normalization by two formulas of pressing,
(2) specify the weight ρ of heterogeneity index I, by homogeney index H and the heterogeneity index I of following formula combined imaging segmentation result image, try to achieve the Image Segmentation evaluation number F (H, I) of each Image Segmentation result images,
F(H,I)=(1-ρ)F(H)+ρF(I);
(3) ask the minimum value of Image Segmentation comprehensive evaluation index F (H, I), segmentation yardstick corresponding to the minimum value of comprehensive evaluation index F (H, I) is the best segmental scale of image.
Embodiment:
The first step, somewhere resolution is selected to be that the QuickBird imagery of 2.4 meters is as experimental data, the eCognition software applying German DEFINIENS company carries out 10 multi-scale division to image, the segmentation yardstick of each segmentation is set to 50 respectively, 60,70 ..., 140, other parameter remains unchanged, and the segmentation result obtained as shown in Figure 1;
Second step, obtains the area a of total number n of the cutting object in each Image Segmentation result images, each cutting object k
kwith gray standard deviation v
k, the homogeney index H of all cutting object inside in each Image Segmentation result images is tried to achieve by following formula,
3rd step, first by formula
try to achieve the mould g (f (x, y)) of the gradient g of raw video f (x, y), then set cutting object bound radius contiguous range radius r=3 in each Image Segmentation result images, if contiguous range is
obtain contiguous range
on number of pixels
by formula
try to achieve contiguous range in each Image Segmentation result images
the mould of upper image gradient | g (f (x, y)) | integration, finally try to achieve the heterogeneity index I in each Image Segmentation result images between imaged object by following formula,
4th step, by homogeney index H and heterogeneity index I normalization, specify the weight ρ of heterogeneity index I, calculate the Image Segmentation comprehensive evaluation index F (H, I) of the Image Segmentation result images under each yardstick, statistics Image Segmentation comprehensive evaluation index F (H, and the relation of segmentation yardstick s I), obtain curve map as shown in Figure 2, the segmentation yardstick that in figure, Image Segmentation comprehensive evaluation index minimum value is corresponding is 130, and namely 130 is the best segmental scale of this image.
Claims (3)
1., based on an image optimum segmentation scale selection method for neighborhood total variation, it is characterized in that comprising the steps:
1) image is split with different segmentation yardsticks, obtain a series of Image Segmentation result images;
2) the homogeney index H of object inside in the Image Segmentation result images under each segmentation yardstick is calculated according to homogeney evaluation method;
3) the heterogeneity index I between object in the Image Segmentation result images under each segmentation yardstick is calculated according to heterogeneous evaluation method;
4) by homogeney index H and heterogeneity index I normalization, specify the weight of heterogeneity index I, calculate the comprehensive evaluation index F (H of the Image Segmentation result images under each yardstick, I), obtain Image Segmentation comprehensive evaluation index F (H, I) minimum value, the segmentation yardstick that the minimum value of comprehensive evaluation index F (H, I) is corresponding is the best segmental scale of image;
Described step 4) comprising:
(1) the homogeney index H of all Image Segmentation result images and heterogeneity index I is distinguished normalization by two formulas of pressing,
(2) specify the weight ρ of heterogeneity index I, by homogeney index H and the heterogeneity index I of following formula combined imaging segmentation result image, try to achieve the Image Segmentation evaluation number F (H, I) of each Image Segmentation result images,
F(H,I)=(1-ρ)F(H)+ρF(I);
(3) ask the minimum value of Image Segmentation comprehensive evaluation index F (H, I), segmentation yardstick corresponding to the minimum value of comprehensive evaluation index F (H, I) is the best segmental scale of image.
2. a kind of segmentation of the image optimum based on neighborhood total variation scale selection method according to claim 1, is characterized in that described step 2) comprising:
(1) total number n of cutting object in each Image Segmentation result images is obtained;
(2) the area a of a kth cutting object in each Image Segmentation result images is set
kwith gray standard deviation v
k;
(3) the homogeney index H of all cutting object inside in each Image Segmentation result images is tried to achieve by following formula,
3. a kind of segmentation of the image optimum based on neighborhood total variation scale selection method according to claim 1, is characterized in that described step 3) comprising:
(1) mould of the gradient g of raw video f (x, y) is tried to achieve by following formula | g (f (x, y)) |,
(2) setting cutting object bound radius contiguous range radius in each Image Segmentation result images is r, and contiguous range is
obtain contiguous range
on number of pixels
(3) contiguous range in each Image Segmentation result images is tried to achieve by following formula
the mould of upper image gradient | g (f (x, y)) | integration,
(4) the heterogeneity index I in each Image Segmentation result images between imaged object is tried to achieve by following formula,
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310386335.7A CN103514599B (en) | 2013-08-30 | 2013-08-30 | A kind of segmentation of the image optimum based on neighborhood total variation scale selection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310386335.7A CN103514599B (en) | 2013-08-30 | 2013-08-30 | A kind of segmentation of the image optimum based on neighborhood total variation scale selection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103514599A CN103514599A (en) | 2014-01-15 |
CN103514599B true CN103514599B (en) | 2016-02-24 |
Family
ID=49897276
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310386335.7A Active CN103514599B (en) | 2013-08-30 | 2013-08-30 | A kind of segmentation of the image optimum based on neighborhood total variation scale selection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103514599B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106529430A (en) * | 2016-10-31 | 2017-03-22 | 武汉大学 | High spatial resolution image residential area extraction method based on characteristics of spatial correlation and heterogeneity |
CN107194942B (en) * | 2017-03-27 | 2020-11-10 | 广州地理研究所 | Method for determining image classification segmentation scale threshold |
CN107067405B (en) * | 2017-03-30 | 2020-04-03 | 河海大学 | Remote sensing image segmentation method based on scale optimization |
CN109816668B (en) * | 2019-01-22 | 2019-12-24 | 中国科学院地理科学与资源研究所 | Remote sensing image unsupervised segmentation evaluation method and device |
CN113850822B (en) * | 2021-09-18 | 2023-04-25 | 四川大学 | Automatic slope unit dividing method based on confluence division |
CN114119645B (en) * | 2021-11-25 | 2022-10-21 | 推想医疗科技股份有限公司 | Method, system, device and medium for determining image segmentation quality |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101126812A (en) * | 2007-09-27 | 2008-02-20 | 武汉大学 | High resolution ratio remote-sensing image division and classification and variety detection integration method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8934698B2 (en) * | 2011-06-22 | 2015-01-13 | The Johns Hopkins University | System and device for characterizing cells |
-
2013
- 2013-08-30 CN CN201310386335.7A patent/CN103514599B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101126812A (en) * | 2007-09-27 | 2008-02-20 | 武汉大学 | High resolution ratio remote-sensing image division and classification and variety detection integration method |
Non-Patent Citations (3)
Title |
---|
付树军 等.基于各向异性扩散方程的超声图像去噪与边缘增强.《电子学报》.2005,第33卷(第7期), * |
何敏 等.面向对象的最优分割尺度计算模型.《大地测量与地球动力学》.2009,第29卷(第1期), * |
李书晓,常红星.基于总变分和形态学的航空图像道路检测算法.《计算机学报》.2007,第30卷(第12期), * |
Also Published As
Publication number | Publication date |
---|---|
CN103514599A (en) | 2014-01-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103514599B (en) | A kind of segmentation of the image optimum based on neighborhood total variation scale selection method | |
CN110378196B (en) | Road visual detection method combining laser point cloud data | |
Zhou et al. | Real-time kiwifruit detection in orchard using deep learning on Android™ smartphones for yield estimation | |
CN106651872B (en) | Pavement crack identification method and system based on Prewitt operator | |
WO2022016884A1 (en) | Method for extracting sea surface wind speed on basis of k-means clustering algorithm | |
CN102254319B (en) | Method for carrying out change detection on multi-level segmented remote sensing image | |
CN112818988B (en) | Automatic identification reading method and system for pointer instrument | |
CN103218605B (en) | A kind of fast human-eye positioning method based on integral projection and rim detection | |
CN104952070B (en) | A kind of corn field remote sensing image segmentation method of class rectangle guiding | |
CN103500449B (en) | Visible remote sensing image cloud detection method of optic on a kind of star | |
CN109141248A (en) | Pig weight measuring method and system based on image | |
CN102110227A (en) | Compound method for classifying multiresolution remote sensing images based on context | |
CN104867150A (en) | Wave band correction change detection method of remote sensing image fuzzy clustering and system thereof | |
CN107909018B (en) | Stable multi-mode remote sensing image matching method and system | |
CN108961330A (en) | The long measuring method of pig body and system based on image | |
CN102855485B (en) | The automatic testing method of one grow wheat heading | |
CN103679701B (en) | Crystal pattern based on Support vector regression is as outline of straight line detection method | |
CN108888294A (en) | A kind of method and device detecting Nuchal translucency width | |
CN110910445B (en) | Object size detection method, device, detection equipment and storage medium | |
CN105354841A (en) | Fast matching method and system for remote sensing images | |
CN108230375A (en) | Visible images and SAR image registration method based on structural similarity fast robust | |
CN104732520A (en) | Cardio-thoracic ratio measuring algorithm and system for chest digital image | |
CN105488541A (en) | Natural feature point identification method based on machine learning in augmented reality system | |
CN103093478A (en) | Different source image rough edge test method based on rapid nuclear spatial fuzzy clustering | |
Liu et al. | Estimating potato above-ground biomass by using integrated unmanned aerial system-based optical, structural, and textural canopy measurements |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |