CN110490088A - DBSCAN Density Clustering method based on region growth method - Google Patents
DBSCAN Density Clustering method based on region growth method Download PDFInfo
- Publication number
- CN110490088A CN110490088A CN201910681442.XA CN201910681442A CN110490088A CN 110490088 A CN110490088 A CN 110490088A CN 201910681442 A CN201910681442 A CN 201910681442A CN 110490088 A CN110490088 A CN 110490088A
- Authority
- CN
- China
- Prior art keywords
- point
- pixel
- dbscan
- eye fundus
- fundus image
- 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
- 238000000034 method Methods 0.000 title claims abstract description 57
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 22
- 230000003287 optical effect Effects 0.000 claims abstract description 19
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000002203 pretreatment Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 abstract description 9
- 238000012545 processing Methods 0.000 abstract description 6
- 230000000694 effects Effects 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 18
- 230000011218 segmentation Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 238000003709 image segmentation Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 4
- 238000010191 image analysis Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 235000001674 Agaricus brunnescens Nutrition 0.000 description 1
- 206010020772 Hypertension Diseases 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 208000030533 eye disease Diseases 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 231100000915 pathological change Toxicity 0.000 description 1
- 230000036285 pathological change Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 210000001210 retinal vessel Anatomy 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/14—Vascular patterns
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Eye Examination Apparatus (AREA)
Abstract
The invention discloses the DBSCAN Density Clustering methods based on region growth method, the present invention space characteristic special according to eye fundus image data set, the relative theory for having used for reference region growth method in Digital Image Processing carries out certain improvement to traditional DBSCAN method, region growth method is first centered on certain point in eye fundus image, with point pixel similar enough in the continuous each central pixel point 8- neighborhood of merger (8 pixels adjacent with the pixel are known as 8- neighborhood), effect that region increases is realized with this.Optical fundus blood vessel extraction is carried out compared to traditional DBSCAN, the time overhead of calculating, which has, to be greatly improved.
Description
Technical field
The DBSCAN method in Density Clustering be the present invention relates to the use of to extract to optical fundus blood vessel, and utilize region
The relative theory of growth method improves tradition DBSCAN method to realize that the acceleration for image data set calculates, and belongs to artificial intelligence
And field of medical image processing.
Background technique
In the epoch of nowadays social economy rapid development, computer and information technology etc. are also and then developed rapidly,
Thus the various new technology caused is dissolved among daily life like the mushrooms after rain, wherein data mining (DM) skill
Art is just born in the continuous research of people.It appear in the 1990s, and this decades it is continuous exploration in, occur
Various data digging methods, wherein clustering is wherein important one of method, and increasingly by
The concern of people.Clustering is also known as cluster analysis, is a kind of important classification method, it be according to the feature of things itself,
A kind of statistical method of Multivariate is carried out to things by established statistical model, its object is to by similar thing
Object is sorted out, these classes are not given in advance, needs to give by clustering to determine type.Clustering is returned
Root knot bottom is to be evolved by classification, however the two is but essentially different, for example classification is needed by training set
Upper training refines some potential rules with this, and for cluster, be in the case where no training sample directly
Data are analyzed, are a kind of unsupervised forms, it can be seen that clustering has its unique existent value and research
Meaning, wherein density clustering method is one of numerous clustering methods classic algorithm, compared to other algorithms
For, Density Clustering proposes the density attributes that data point itself possessed on the basis of distance, and then is clustered with this, together
Traditional clustering method compares, the cluster being more suitable under spatial data, but due to being related to the distance at arbitrary number strong point
It calculates, so the time complexity of DBSCAN algorithm has often reached the rank of O (n^2), the time calculated for mass data
Expense is often unsatisfactory.
With keeping updating for electronic equipment, people in daily life for the dependence of these equipment increasingly
By force, the mean vision level of people is also declining year by year at the same time, therefore and the number of suffering from eye disease increase year by year
Add, China is most one of the country of current blind person's number according to statistics, there are about 6,700,000 people, when the data volume of such a terror
Quarter warns us more to focus on the protection of eyes, and medically, can anti-Zhiduo County in advance by the analysis of eye fundus image
Kind blinding disease.Eye fundus image is a kind of diagnosis imaging obtained in ophthalmology by fundus camera, is divided eye fundus image
Analysis processing can be found that the variation of blood vessel structure, thus for the diagnosing and treating of eyeground pathological changes caused by diabetes, hypertension etc.
Auxiliary information is provided, there is more important realistic meaning for the blood vessel segmentation in eye fundus image thus.It relates among these
The image Segmentation Technology in Digital Image Processing has been arrived, the research of image segmentation is constantly subjected to the great attention of people for many years,
So far various types of partitioning algorithms are proposed, wherein clustering algorithm is a kind of wherein important method, and image Segmentation Technology is just
It is the technology and process for dividing the image into several regions specific, with unique properties and proposing interesting target.It is
By the committed step of image procossing to image analysis, the accuracy of image segmentation directly affects subsequent image under normal circumstances
The accuracy of analysis, thus a good image segmentation algorithm becomes particularly important.In this way especially for optical fundus blood vessel segmentation
Application in professional domain, even more extremely harsh for the various requirement of image procossing, how efficiently and accurately to divide thus
Cut the emphasis that blood vessel is increasingly becoming people's research.
Summary of the invention
In the application that optical fundus blood vessel extracts, training is used as due to being difficult to obtain segmentation result with high accuracy in practice
The label training of collection, such as have in the database D RIVE of eyeground the manual segmentation result of two experts, made with one of
There was only 95% or so to compare the accuracy rate of another segmentation result, it can be seen that very high for label required precision
It is not very applicable for having supervision class deep learning algorithm, and for traditional unsupervised class machine learning method, it does not need to instruct
Practice data, is more appropriately applied in the scene of optical fundus blood vessel extraction.
In traditional Unsupervised clustering algorithm, the advantages of DBSCAN method based on Density Clustering be this method for
Noise is insensitive and can find clustering for any shape, is relatively suitble to optical fundus blood vessel to extract such background data more complicated
In scene, while the disadvantages of this method is that time complexity is relatively high, has reached O (n^2) rank, thus the bring time opens
Sell larger, therefore the present invention uses for reference the relative theory of region growth method in image procossing on the basis of traditional DBSCAN method
Certain improvement is carried out to DBSCAN method, so that the time complexity of this method drops to O (n) rank, sharp fall
The time overhead of calculating.
The purpose of traditional DBSCAN algorithm is to find the high-density region separated by density regions.The core concept of algorithm
It is to find the higher point of density first, then the similar high density point near the point joins together to generate various cluster.
The realization of the algorithm, which mainly passes through, is arranged two important parameters, i.e. radius and density threshold (Eps and MinPts), and according to
The two parameters divide data points all in eye fundus image data set, first by x, y of pixel each in eye fundus image
Coordinate and pixel value as a ternary to the data point being expressed as in a three-dimensional space, next will according to following rule
These data points are divided into following three categories:
1) core point: the Eps neighborhood for a certain data point (is center eps as radius to put under three-dimensional space
Spheric neighbo(u)rhood) it includes and has more than MinPts number point of destination.
2) marginal point: being unsatisfactory for core point requirement, but falls in the point in certain core point Eps neighborhood.
3) noise spot: neither core point is also not the point of marginal point.
There is above-mentioned concept, can be described as using the process that traditional DBSCAN algorithm carries out optical fundus blood vessel extraction:
1. inputting eye fundus image
2. carrying out the pretreatments such as mean filter, contrast enhancing to image
3. whole pixels are organized into three-dimensional data points according to sequence from left to right, from top to bottom
4. a untreated point P is found at each Three-ary Number strong point of sequential processes
5. whether calculating in the Eps neighborhood of point P comprising being greater than a data point of MinPts, turn if being unsatisfactory for the requirement
9. turning 6. if meeting the requirement
6. P is labeled as core point, a new cluster C is established, is turned 7..
7. all core points for being connected to (including the data point in Eps neighborhood) with point P and marginal point are all added to C
In, turn 9..
8. label P is noise spot, turn 9..
9. turning 4., otherwise turning 10. if there are also unlabelled data points in data set.
10. end of clustering exports result.
It just completes the optical fundus blood vessel based on DBSCAN according to the flow operations to extract, the flow chart of entire algorithm such as Fig. 1
It is shown.
The present invention space characteristic special according to eye fundus image data set, has used for reference region growth method in Digital Image Processing
Relative theory certain improvement carried out to traditional DBSCAN method, during region growth method is with certain point in eye fundus image first
The heart, the continuous each central pixel point 8- neighborhood of merger (8 pixels adjacent with the pixel are known as 8- neighborhood) is interior and should
The similar enough pixel of point realizes effect that region increases with this.It will be right in traditional DBSCAN method according to this thought
The decision condition of three classes pixel is modified as follow in eye fundus image:
1) core point: it is less than the number of Eps point in some pixel 8- neighborhood in eye fundus image with pixel distance
Mesh is greater than the pixel of minimum number MinPts.
2) it is unsatisfactory for core point requirement, but falls in 8 neighborhood of core point the point for belonging to marginal point.
3) neither core point is also not the point of marginal point.
According to modified decision condition, the DBSCAN optical fundus blood vessel extracting method process based on region growth method can be retouched
It states are as follows:
S1 inputs eye fundus image；
S2 carries out the pretreatments such as mean filter, contrast enhancing to input eye fundus image；
S3 finds a untreated point P according to sequence from left to right, from top to bottom to pretreated eye fundus image；
Whether S4 calculates comprising being less than Eps greater than MinPts pixel and the point P margin of image element in 8 neighborhoods of point P,
Turn S8 if being unsatisfactory for the requirement, if meeting the requirement, turns S5；
Point P is labeled as core point by S5, is established a new cluster C, is turned S6.
All core points that (including the data point in Eps neighborhood) is connected to point P and marginal point are all added to newly by S6
In cluster C, turn S8.
S7 mark point P is noise spot, turns S8.
If S8 is inputted in the data set of the acquired pixel of eye fundus image, there are also unlabelled data points, turn S3, otherwise
Turn S9.
S9 end of clustering exports result.
The DBSCAN optical fundus blood vessel extracting method based on region growth method is just completed according to the flow operations, it is improved
Algorithm flow chart is as shown in Figure 2.
Optical fundus blood vessel extraction is carried out using traditional DBSCAN method, is related to finding out in whole pixels and point P distance
Pixel less than Eps, will calculate in whole pixels that every bit is at a distance from P in general, the time of entire algorithm
Complexity is to have reached O (n^2) rank；And optical fundus blood vessel extraction is carried out using improved DBSCAN method, for each picture
Vegetarian refreshments P is only focused in 8 pixels in 8 neighborhood of P point, so the processing time of each pixel is constant rank,
The time complexity of entire algorithm is O (n) rank, it can be seen that, optical fundus blood vessel extraction is carried out compared to tradition DBSCAN, calculating
Time overhead, which has, to be greatly improved.
Detailed description of the invention
Fig. 1 is to extract optical fundus blood vessel flow chart using tradition DBSCAN method
Fig. 2 is to extract optical fundus blood vessel flow chart using improvement DBSCAN method
Fig. 3 is the result being applied to algorithm in optical fundus blood vessel extraction.
Specific embodiment
In order to verify it is proposed by the present invention based on the DBSCAN method of region growth method optical fundus blood vessel extract application in
Performance, compares with traditional DBSCAN method.It is the displaying of experimental result as shown in Figure 3.Tradition is utilized in document [1]
DBSCAN method carries out the extraction of optical fundus blood vessel to the eyeground DRIVE database, and sensitivity and cost time are respectively 73.82%
And 83s, and the DBSCAN Density Clustering method based on region growth method carries out the sensitivity of vessel extraction and the cost time is
81.69% and 12s, it can be seen that improved method has significantly in segmentation accuracy rate and on time overhead
It improves.
Citation
[1]Shiraz,Iran.Retinal Vessel Segmentation Using System Fuzzy and
DBSCAN Algorithm,2015 2nd International Conference on Pattern Recognition and
Image Analysis (IPRIA 2015)March 11-12,2015。
Claims (2)
1. the DBSCAN Density Clustering method based on region growth method, it is characterised in that: this method comprises the following steps,
S1 inputs eye fundus image；
S2 carries out the pretreatments such as mean filter, contrast enhancing to input eye fundus image；
S3 finds a untreated point P according to sequence from left to right, from top to bottom to pretreated eye fundus image；
Whether S4 calculates in 8 neighborhoods of point P comprising being greater than MinPts pixel and the point P margin of image element less than Eps, if discontented
The foot requirement then turns S8, if meeting the requirement, turns S5；
Point P is labeled as core point by S5, is established a new cluster C, is turned S6；
S6 all core points for including data point in Eps neighborhood will be connected to point P and marginal point is all added in new cluster C,
Turn S8；
S7 mark point P is noise spot, turns S8；
If S8 is inputted in the data set of the acquired pixel of eye fundus image, there are also unlabelled data points, turn S3, otherwise turn S9；
S9 end of clustering exports result；
The DBSCAN optical fundus blood vessel extracting method based on region growth method is just completed according to the flow operations.
2. the DBSCAN Density Clustering method according to claim 1 based on region growth method, it is characterised in that:
The decision condition of three classes pixel in eye fundus image is modified as follow,
1) core point: the number for being less than Eps point with pixel distance in some pixel 8- neighborhood in eye fundus image is big
In the pixel of minimum number MinPts；
2) it is unsatisfactory for core point requirement, but falls in 8 neighborhood of core point the point for belonging to marginal point；
3) neither core point is also not the point of marginal point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910681442.XA CN110490088A (en) | 2019-07-26 | 2019-07-26 | DBSCAN Density Clustering method based on region growth method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910681442.XA CN110490088A (en) | 2019-07-26 | 2019-07-26 | DBSCAN Density Clustering method based on region growth method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110490088A true CN110490088A (en) | 2019-11-22 |
Family
ID=68548523
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910681442.XA Pending CN110490088A (en) | 2019-07-26 | 2019-07-26 | DBSCAN Density Clustering method based on region growth method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110490088A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112070140A (en) * | 2020-09-01 | 2020-12-11 | 中国人民解放军陆军工程大学 | Density clustering mark-like pattern recognition method based on dimension decomposition |
| CN113484875A (en) * | 2021-07-30 | 2021-10-08 | 燕山大学 | Laser radar point cloud target hierarchical identification method based on mixed Gaussian ordering |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101286199A (en) * | 2007-09-14 | 2008-10-15 | 西北工业大学 | Method of image segmentation based on area upgrowth and ant colony clustering |
-
2019
- 2019-07-26 CN CN201910681442.XA patent/CN110490088A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101286199A (en) * | 2007-09-14 | 2008-10-15 | 西北工业大学 | Method of image segmentation based on area upgrowth and ant colony clustering |
Non-Patent Citations (2)
| Title |
|---|
| 宝G: ""DBSCAN聚类算法"", 《HTTPS://BLOG.CSDN.NET/U013611461/ARTICLE/DETAILS/53639835》 * |
| 郭世可: ""一种结合密度聚类和区域生长的图像分割方法"", 《计算机研究与发展》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112070140A (en) * | 2020-09-01 | 2020-12-11 | 中国人民解放军陆军工程大学 | Density clustering mark-like pattern recognition method based on dimension decomposition |
| CN112070140B (en) * | 2020-09-01 | 2022-05-03 | 中国人民解放军陆军工程大学 | Density clustering mark-like pattern recognition method based on dimension decomposition |
| CN113484875A (en) * | 2021-07-30 | 2021-10-08 | 燕山大学 | Laser radar point cloud target hierarchical identification method based on mixed Gaussian ordering |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106408562B (en) | Eye fundus image Segmentation Method of Retinal Blood Vessels and system based on deep learning | |
| CN106650806B (en) | A kind of cooperating type depth net model methodology for pedestrian detection | |
| CN112132817B (en) | Retina blood vessel segmentation method for fundus image based on mixed attention mechanism | |
| CN109376636A (en) | Eye ground image classification method based on capsule network | |
| CN107977671A (en) | A kind of tongue picture sorting technique based on multitask convolutional neural networks | |
| CN108346145A (en) | The recognition methods of unconventional cell in a kind of pathological section | |
| CN107506761A (en) | Brain image dividing method and system based on notable inquiry learning convolutional neural networks | |
| CN108776774A (en) | A kind of human facial expression recognition method based on complexity categorization of perception algorithm | |
| CN103366180A (en) | Cell image segmentation method based on automatic feature learning | |
| CN111539320B (en) | Multi-view gait recognition method and system based on mutual learning network strategy | |
| CN115205300A (en) | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion | |
| KR101687217B1 (en) | Robust face recognition pattern classifying method using interval type-2 rbf neural networks based on cencus transform method and system for executing the same | |
| CN109886161A (en) | A kind of road traffic index identification method based on possibility cluster and convolutional neural networks | |
| CN112884788B (en) | Cup optic disk segmentation method and imaging method based on rich context network | |
| CN111008647B (en) | Sample extraction and image classification method based on void convolution and residual linkage | |
| Chen et al. | Spherical-patches extraction for deep-learning-based critical points detection in 3D neuron microscopy images | |
| CN113269103A (en) | Abnormal behavior detection method and system based on space map convolutional network | |
| CN110490088A (en) | DBSCAN Density Clustering method based on region growth method | |
| Reddy et al. | Novel algorithm based on region growing method for better image segmentation | |
| CN112669343A (en) | Zhuang minority nationality clothing segmentation method based on deep learning | |
| CN109872299A (en) | A kind of Chinese medicine tongue color coating colour recognition methods | |
| Ma et al. | Retinal vessel segmentation based on generative adversarial network and dilated convolution | |
| CN109165551B (en) | Expression recognition method for adaptively weighting and fusing significance structure tensor and LBP characteristics | |
| CN101140660B (en) | Backbone pruning method based on discrete curve evolvement | |
| CN104200201B (en) | A kind of coding method and system applied to iris recognition |
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: 20191122 |
|
| RJ01 | Rejection of invention patent application after publication |