CN112991361A - Image segmentation method based on local graph structure similarity - Google Patents

Image segmentation method based on local graph structure similarity Download PDF

Info

Publication number
CN112991361A
CN112991361A CN202110263189.3A CN202110263189A CN112991361A CN 112991361 A CN112991361 A CN 112991361A CN 202110263189 A CN202110263189 A CN 202110263189A CN 112991361 A CN112991361 A CN 112991361A
Authority
CN
China
Prior art keywords
image
norm
matrix
sub
average
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110263189.3A
Other languages
Chinese (zh)
Other versions
CN112991361B (en
Inventor
罗西玲
陈晨
康蕊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Laser and Optoelectronics Intelligent Manufacturing of Wenzhou University
Original Assignee
Institute of Laser and Optoelectronics Intelligent Manufacturing of Wenzhou University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Institute of Laser and Optoelectronics Intelligent Manufacturing of Wenzhou University filed Critical Institute of Laser and Optoelectronics Intelligent Manufacturing of Wenzhou University
Priority to CN202110263189.3A priority Critical patent/CN112991361B/en
Publication of CN112991361A publication Critical patent/CN112991361A/en
Application granted granted Critical
Publication of CN112991361B publication Critical patent/CN112991361B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a self-adaptive image segmentation method based on local graph structure similarity, which comprises the following steps of uniformly dividing an original image into a plurality of non-overlapping sub-images and calculating the average sub-image; then, establishing an average subgraph image connection matrix according to the graph theory and calculating the eigenvalue and the eigenvector of the average subgraph image connection matrix; secondly, sequentially selecting each subinterval of the original image and decomposing the subinterval by using the obtained characteristic vector; then decomposing the decomposed matrix into a diagonal matrix and a non-diagonal matrix and calculating the 1 norm of the non-diagonal matrix; after the 1 norm of all the subintervals is calculated, performing linearization processing on the 1 norm of all the subintervals; and finally, carrying out image segmentation according to the norm distribution. The invention has the characteristics of strong anti-interference capability and high robustness.

Description

Image segmentation method based on local graph structure similarity
Technical Field
The invention relates to the technical field of image processing, in particular to an image segmentation method based on local graph structure similarity.
Background
Image segmentation is a challenging task in the field of digital image processing. The accurate segmentation of the image has important research significance and application value in the aspects of national defense safety, medical images, industrial measurement, automatic driving and the like. The traditional image segmentation method is to establish a segmentation model by using a mathematical method on the basis of a feature space, and the common traditional methods comprise an edge-based segmentation method, a threshold-based segmentation method, a region-based segmentation method and the like. In the segmentation process, feature expression is usually manually designed, most of the image expression is low-level features, and the image expression is greatly influenced by factors such as discontinuous image boundary information, less application of image spatial relationship and the like. In addition, the traditional detection algorithm is sensitive to environmental interference such as illumination change, shadow shielding and the like, so that the problem of insufficient robustness and the like is solved, and the traditional detection algorithm cannot be directly applied to engineering scenes. With the increase of digital image processing and computing power, data representation methods based on graph structures have received a great deal of attention. The invention provides an image segmentation method based on local graph structure similarity, and is expected to be used for image region segmentation.
Disclosure of Invention
The invention aims to provide an image segmentation method based on local graph structure similarity. The method has the characteristic of strong anti-noise interference capability, and can perform region segmentation on the image by utilizing the graph structure.
The technical scheme of the invention is as follows: an image segmentation method based on local graph structure similarity comprises the following steps:
the method comprises the following steps:
s1: uniformly dividing an original image into a plurality of non-overlapping sub-images and calculating the average sub-image;
s2: establishing an average subgraph connection matrix;
s3: calculating a feature vector of the average subgraph connection matrix;
s4: setting a sliding window, wherein the size of the sliding window is the same as the size of the sub-image;
s5: moving the sliding window to the upper left corner of the original image, and establishing an image connection matrix of pixel point gray values in the current window;
s6: decomposing the graph connection matrix obtained by the current window by using the obtained eigenvector to obtain matrix information after the current window is decomposed;
s7: decomposing the decomposed matrix information into a diagonal matrix and a non-diagonal matrix, and calculating 1 norm of the non-diagonal matrix;
s8: moving the sliding window one pixel to the right, repeating steps S6-S7;
s9: continuously traversing all pixels of the current picture to form a 1 norm sequence of the non-diagonal matrix of the sliding window;
s10: carrying out normalization processing on the obtained 1 norm sequence;
s11: and setting an input threshold, wherein if the 1 norm is greater than the threshold, the pixel points corresponding to the norm belong to the same category, and otherwise, the pixel points belong to another category.
In the above adaptive image segmentation method based on the similarity of local graph structures, in step S2, the algorithm for establishing the average sub-graph connection matrix is as follows:
Figure BDA0002970952900000031
wherein X is the established average sub-image connection matrix, d is the Euclidean distance between related pixel points, subscript is the corresponding serial number of the sub-image, and for the sub-image containing n pixel points, the total elements of the average connection matrix are n multiplied by n.
In the foregoing adaptive image segmentation method based on local graph structure similarity, in step S6, an algorithm for decomposing a current sliding window element by using an obtained feature vector is as follows:
Yt＝ΓXtΓ'
wherein, XtAs current sliding window element, YtΓ is the eigenvector calculated in step S3 for the matrix information decomposed by the current sliding window element, and t is the local window sequence number.
In the foregoing adaptive image segmentation method based on local graph structure similarity, in step S7, the algorithm of the normalization process is:
Figure BDA0002970952900000032
wherein, N is 1 norm after local window element calculation, N' is 1 norm after normalization, NminIs the minimum of the original 1-norm sequence, NmaxIs the maximum of the original 1-norm sequence.
Compared with the prior art, the invention has the following beneficial effects:
firstly, uniformly dividing an original image into a plurality of sub-images which are not overlapped with each other and calculating the average sub-image; then, establishing an average subgraph image connection matrix according to the graph theory and calculating the feature vector of the average subgraph image connection matrix; traversing all sliding windows by using the obtained feature vectors and decomposing elements of the sliding windows by using the obtained feature vectors; then decomposing the decomposed matrix into a diagonal matrix and a non-diagonal matrix and calculating the 1 norm of the non-diagonal matrix; after the 1 norm of all the subintervals is calculated, carrying out normalized processing on the 1 norm of all the subintervals; and finally, carrying out image segmentation on the normalized 1 norm according to an input threshold. The method uses the structure similarity of the local graph to segment the image, and has stronger anti-noise interference capability.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a photograph of a crack in a road surface used in example 1 of the present invention;
FIG. 3 is a schematic view of the segmentation of a sub-image of a pavement crack photograph in example 1 of the present invention;
fig. 4 is a schematic diagram of a sub-image extracted in embodiment 1 of the present invention, in which the window side length is 3;
fig. 5 is a schematic diagram of an euclidean distance calculated by taking a certain pixel point as a reference in a neutron image in embodiment 1 of the present invention;
FIG. 6 is a graph connection matrix generated in embodiment 1 of the present invention;
FIG. 7 is a feature vector generated in embodiment 1 of the present invention;
fig. 8 shows the result of image segmentation in example 1 of the present invention, in which the threshold is 3 times of the average value of the generated 1-norm sequence.
Detailed Description
The invention is further illustrated by the following figures and examples, which are not to be construed as limiting the invention.
Example 1: the invention is further explained below by combining specific cases, and a flow chart thereof is shown in fig. 1, and includes the following steps,
1) as shown in fig. 2 and fig. 3, the obtained pavement crack photo is uniformly divided into a plurality of non-overlapping sub-images, and the average sub-image is calculated;
2) as shown in fig. 4, local matrix information (shown on the right side of fig. 4) can be obtained by extracting the gray-scale values (P1 to P9) of the pixels in the sliding window;
3) as shown in fig. 5, by calculating euclidean distances between the pixels (e.g., the distances between the P1 point and P2, P3 …, P9 in fig. 5), the distance relationship between the pixels can be obtained;
4) as shown in fig. 6, an average sub-image connection matrix may be established according to the graph theory based on the distance relationship between the pixels (as shown in the matrix information of fig. 6, the first row is the distance between the point P1 in fig. 5 and P2, P3 …, P9, respectively), where the establishment algorithm of the average sub-image connection matrix is:
Figure BDA0002970952900000051
wherein X is the established average sub-image connection matrix, d is the Euclidean distance between related pixel points, subscript is the corresponding serial number of the sub-image, and for the sub-image containing n pixel points, the total elements of the average connection matrix are n multiplied by n;
5) calculating the feature vector of the average subgraph connection matrix, wherein the feature vector is shown in FIG. 7 in the example;
6) setting a sliding window, wherein the size of the sliding window is the same as the size of the sub-image, moving the sliding window to the upper left corner of the original image, and establishing a graph connection matrix of the gray value of the pixel point in the current window; traversing the sliding window by using the obtained feature vector, and decomposing the sliding window by using the obtained feature vector, wherein an algorithm for decomposing the local window by using the obtained feature vector is as follows:
Yt＝ΓXtΓ'
wherein, XtFor the current local window element, YtΓ is the eigenvector calculated in step S3 for the matrix information after the local window element decomposition, and t is the local window sequence number.
7) Decomposing the decomposed matrix into a diagonal matrix and a non-diagonal matrix;
8) calculating the 1 norm of the non-diagonal matrix, wherein the algorithm is as follows:
Figure BDA0002970952900000061
wherein M isnon-diagIs a matrix off-diagonal element sequence, | | | | | non-conducting phosphor1Calculating the sign for a 1 norm, Mnon-diagIs the ith element of the matrix off-diagonal element sequence, and n is the length of the matrix off-diagonal element sequence;
9) moving the sliding window one pixel to the right in sequence, so as to decompose all local windows, calculate 1 norm of the local windows to form a 1 norm sequence, and carry out standardization processing on the 1 norm sequence, wherein the used norm sequence standardization processing algorithm is as follows:
Figure BDA0002970952900000062
wherein N is a 1-norm sequence after local window element calculation, N' is a normalized 1-norm sequence, and NminIs the minimum of the original 1-norm sequence, NmaxIs the maximum value of the original 1 norm sequence;
8) and setting an input threshold, carrying out image segmentation on the normalized 1 norm according to the input threshold, wherein if the 1 norm is greater than the threshold, the pixel point (the central point position of the sliding serial port) corresponding to the norm is in the same category, and otherwise, the pixel point is in the other category. In the present example, the threshold is set to 3 times the average value of the generated 1-norm series, and the result is shown in fig. 8. As shown in fig. 8, under the background noise interference (the left part of fig. 8 has obvious illumination shadows), the image segmentation method provided by the invention can still better realize the image segmentation effect, which indicates that the image segmentation method with local graph structure similarity provided by the invention has stronger anti-noise interference capability.

Claims (4)

1. A self-adaptive image segmentation method based on local graph structure similarity is characterized in that: the method comprises the following steps:
s1: uniformly dividing an original image into a plurality of non-overlapping sub-images and calculating the average sub-image;
s2: establishing an average subgraph connection matrix;
s3: calculating a feature vector of the average subgraph connection matrix;
s4: setting a sliding window, wherein the size of the sliding window is the same as the size of the sub-image;
s5: moving the sliding window to the upper left corner of the original image, and establishing an image connection matrix of pixel point gray values in the current window;
s6: decomposing the graph connection matrix obtained by the current window by using the obtained eigenvector to obtain matrix information after the current window is decomposed;
s7: decomposing the decomposed matrix information into a diagonal matrix and a non-diagonal matrix, and calculating 1 norm of the non-diagonal matrix;
s8: moving the sliding window one pixel to the right, repeating steps S6-S7;
s9: continuously traversing all pixels of the current picture to form a 1 norm sequence of the partial window non-diagonal matrix;
s10: carrying out normalization processing on the obtained 1 norm sequence;
s11: and setting an input threshold, wherein if the 1 norm is greater than the threshold, the pixel points corresponding to the norm belong to the same category, and otherwise, the pixel points belong to another category.
2. The adaptive image segmentation method based on the similarity of local graph structures according to claim 1, characterized in that: in step S2, the algorithm for establishing the average sub-image connection matrix is:
Figure FDA0002970952890000021
wherein X is the established average sub-image connection matrix, d is the Euclidean distance between related pixel points, subscript is the corresponding serial number of the sub-image, and for the sub-image containing n pixel points, the total elements of the average connection matrix are n multiplied by n.
3. The adaptive image segmentation method based on the similarity of local graph structures according to claim 1, characterized in that: in step S6, the algorithm for decomposing the local window using the obtained feature vector is as follows:
Yt＝ΓXtΓ'
wherein, XtAs current sliding window element, YtΓ is the eigenvector obtained in step S3 for the matrix information decomposed by the current sliding window element, and t is the sliding window sequence number.
4. The adaptive image segmentation method based on the similarity of local graph structures according to claim 1, characterized in that: in step S10, the algorithm of the normalization process is:
Figure FDA0002970952890000022
wherein, N is a 1 norm sequence after the local window element is calculated, N' is a normalized 1 norm sequence, and NminIs the minimum of the original 1-norm sequence, NmaxIs the maximum of the original 1-norm sequence.
CN202110263189.3A 2021-03-11 2021-03-11 Image segmentation method based on local graph structure similarity Active CN112991361B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110263189.3A CN112991361B (en) 2021-03-11 2021-03-11 Image segmentation method based on local graph structure similarity

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110263189.3A CN112991361B (en) 2021-03-11 2021-03-11 Image segmentation method based on local graph structure similarity

Publications (2)

Publication Number Publication Date
CN112991361A true CN112991361A (en) 2021-06-18
CN112991361B CN112991361B (en) 2023-06-13

Family

ID=76334974

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110263189.3A Active CN112991361B (en) 2021-03-11 2021-03-11 Image segmentation method based on local graph structure similarity

Country Status (1)

Country Link
CN (1) CN112991361B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117746260A (en) * 2024-02-20 2024-03-22 山东省国土测绘院 Remote sensing data intelligent analysis method and system

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090028433A1 (en) * 2007-05-03 2009-01-29 David Allen Tolliver Method for partitioning combinatorial graphs
CN102867307A (en) * 2012-09-10 2013-01-09 西安电子科技大学 SAR image segmentation method based on feature vector integration spectral clustering
CN106023221A (en) * 2016-05-27 2016-10-12 哈尔滨工业大学 Remote sensing image segmentation method based on nonnegative low-rank sparse correlated drawing
US20180174301A1 (en) * 2016-12-20 2018-06-21 Kodak Alaris, Inc. Iterative method for salient foreground detection and multi-object segmentation
CN108876803A (en) * 2018-05-25 2018-11-23 西安理工大学 A kind of color image segmentation method based on spectral clustering community division
CN109214428A (en) * 2018-08-13 2019-01-15 平安科技（深圳）有限公司 Image partition method, device, computer equipment and computer storage medium
CN110020599A (en) * 2019-02-28 2019-07-16 浙江工业大学 A kind of facial image clustering method of sparse enhanced type low-rank constraint
CN110599461A (en) * 2019-08-21 2019-12-20 东南大学 Thalamus function partitioning method based on subspace feature learning

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090028433A1 (en) * 2007-05-03 2009-01-29 David Allen Tolliver Method for partitioning combinatorial graphs
CN102867307A (en) * 2012-09-10 2013-01-09 西安电子科技大学 SAR image segmentation method based on feature vector integration spectral clustering
CN106023221A (en) * 2016-05-27 2016-10-12 哈尔滨工业大学 Remote sensing image segmentation method based on nonnegative low-rank sparse correlated drawing
US20180174301A1 (en) * 2016-12-20 2018-06-21 Kodak Alaris, Inc. Iterative method for salient foreground detection and multi-object segmentation
CN108876803A (en) * 2018-05-25 2018-11-23 西安理工大学 A kind of color image segmentation method based on spectral clustering community division
CN109214428A (en) * 2018-08-13 2019-01-15 平安科技（深圳）有限公司 Image partition method, device, computer equipment and computer storage medium
CN110020599A (en) * 2019-02-28 2019-07-16 浙江工业大学 A kind of facial image clustering method of sparse enhanced type low-rank constraint
CN110599461A (en) * 2019-08-21 2019-12-20 东南大学 Thalamus function partitioning method based on subspace feature learning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
PEDRO F. FELZENSZWALB ET AL: "Efficient Graph-Based Image Segmentation", 《INTERNATIONAL JOURNAL OF COMPUTER VISION》 *
龚楷椿等: "基于邻接矩阵的自适应图像分割算法研究", 《电子科技》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117746260A (en) * 2024-02-20 2024-03-22 山东省国土测绘院 Remote sensing data intelligent analysis method and system
CN117746260B (en) * 2024-02-20 2024-05-31 山东省国土测绘院 Remote sensing data intelligent analysis method and system

Also Published As

Publication number Publication date
CN112991361B (en) 2023-06-13

Similar Documents

Publication Publication Date Title
CN108108746B (en) License plate character recognition method based on Caffe deep learning framework
WO2022199583A1 (en) Image processing method and apparatus, computer device, and storage medium
WO2021008019A1 (en) Posture tracking method and apparatus, and computer-readable storage medium
CN107038416B (en) Pedestrian detection method based on binary image improved HOG characteristics
US20140093163A1 (en) Perceptually lossless color compression
CN111126115B (en) Violent sorting behavior identification method and device
CN105590319A (en) Method for detecting image saliency region for deep learning
CN108345835B (en) Target identification method based on compound eye imitation perception
CN114359323A (en) Image target area detection method based on visual attention mechanism
Han et al. An enhanced image binarization method incorporating with Monte-Carlo simulation
Zhang et al. Single image dehazing based on bright channel prior model and saliency analysis strategy
CN112991361B (en) Image segmentation method based on local graph structure similarity
CN112330639A (en) Significance detection method for color-thermal infrared image
Fang et al. 1-D barcode localization in complex background
CN110378307B (en) Texture image direction field estimation method based on deep learning
CN110349119B (en) Pavement disease detection method and device based on edge detection neural network
CN110930358A (en) Solar panel image processing method based on self-adaptive algorithm
Mikhov et al. Fuzzy logic approach to improving the digital images contrast
Chen et al. Image segmentation based on Bethe approximation for Gaussian mixture model
Krupiński et al. Improved two-step binarization of degraded document images based on Gaussian mixture model
CN105139428A (en) Quaternion based speeded up robust features (SURF) description method and system for color image
Jung et al. Text extraction in real scene images on planar planes
CN110781977A (en) Motion shadow detection method and system based on spatial correlation and extreme learning machine
CN114359244B (en) Image saliency detection method based on super-pixel segmentation and multiple color features
CN114677568B (en) Linear target detection method, module and system based on neural network

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
GR01 Patent grant
GR01 Patent grant