CN106157279A - Eye fundus image lesion detection method based on morphological segment - Google Patents
Eye fundus image lesion detection method based on morphological segment Download PDFInfo
- Publication number
- CN106157279A CN106157279A CN201510126237.9A CN201510126237A CN106157279A CN 106157279 A CN106157279 A CN 106157279A CN 201510126237 A CN201510126237 A CN 201510126237A CN 106157279 A CN106157279 A CN 106157279A
- Authority
- CN
- China
- Prior art keywords
- image
- region
- eye fundus
- fundus image
- hemorrhage
- 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
Landscapes
- Eye Examination Apparatus (AREA)
Abstract
A kind of eye fundus image lesion detection method based on morphological segment, first carries out smothing filtering, and uses region-growing method to position optic disc region eye fundus image；Then obtain removing the optical fundus background image of seepage areas by Morphological scale-space, and obtain including blood vessel and hemorrhage Mixed Zone image by Threshold segmentation；Carrying out rim detection finally by Kirsch operator and obtain blood-vessel image by region growing, it is hemorrhagic areas image with the difference of Mixed Zone image and is calculated this region area.Present invention, avoiding the interference that the blood vessel of similar gray value may bring in other hemorrhage detection algorithm, accelerate detection speed the most simultaneously.
Description
Technical field
The present invention relates to the technology of a kind of image processing field, specifically a kind of based on eye fundus image healthy area background reduction
Hard exudate detection method and based on rim detection and the hemorrhage detection method of region growing.
Background technology
Along with the development of computer graphics disposal technology, the analysis to eye fundus image no longer relies solely on the naked eyes sight in ophthalmologist
Examining, as far back as the seventies and eighties in 20th century, foreign scholar just has been proposed for based on eye fundus image process rigid with the detection automatically analyzed
Ooze out and hemorrhage technology, and carried out substantial amounts of research.Traditional artificial qualitative analysis lacks quantization means.Dependence computer is fast
Speed identifies the focus in eye fundus image the most automatically, can avoid doctor that patient injection medicaments makes its eye fundus image clear,
The artificial diagosis that simultaneously it also avoid doctor judges, saves a large amount of manpower and materials and time, for the enforcement of extensive examination pathological changes
Provide the foundation condition.Its achievement in research has great realistic meaning in field of medical image processing.
At present conventional eye fundus image detection is oozed out has a morphological operation with hemorrhage method, Threshold segmentation, algorithm of region growing,
SVM classifier, cluster analysis based on Markov model.Morphological operation is mainly by form that is hemorrhage and that ooze out with blood vessel not
Identical, eliminate noise after image is carried out opening and closing operations, then by Morphology observation, blood vessel is split away from image, stay
Hard exudate and hemorrhage.After thresholding method is mainly by being changed into gray-scale map by eye fundus image, hemorrhage gray scale is relatively low,
The gray scale of hard exudate is of a relatively high, by using appropriate threshold value segmentation image to detect hemorrhage with oozing out.Algorithm of region growing
Taking suitable seed points by image being carried out sampling, then carrying out region growing, the pixel of similar gray value is developed into bigger district
Territory, includes with oozing out hemorrhage.SVM classifier and cluster analysis based on Markov model are by entering substantial amounts of data
Row sample analysis, enable a computer to identify by the way of machine learning which place be which place hemorrhage be to ooze out.With
Under be the analysis and evaluation of all kinds of method.
Table 1 method for processing fundus images
Generally speaking, existing method for processing fundus images is primarily present following defect:
(1) precision is low, it is impossible to carry out accurate quantitative analysis.
(2) needing fixing parameter combination, universality is little, it is impossible to analyze accurately for multiple image.
(3) need mass data as training sample.For not having mass data can not have effect as the user of sample.
(4) some algorithm calculation cost is excessive, calculates cost the highest.Computing capability beyond general computer.
Although there being the achievement in research analyzed about eye fundus image in a large number at present, but the algorithm that expense is the biggest, need mass data to make
Based on machine learning, and precision is poor, the additive method that universality is the highest can not meet current hospital to eye fundus image at
The demand of reason.Therefore, need a kind of Computer Image Processing rapidly and efficiently badly and be applied to facing of diabetic ophthalmopathy with PRS
Bed diagnosis.
Through finding the literature search of prior art, Kande GB, Savithri TS et al. was at " IEEE Int.Symp. in 2009
Biomed.Imaging:From Nano to Macro " " the Detection of red lesions that delivers on page 558 to page 561
In digital fundus images " detection of lesion region (in the digital fundus image red) article proposes a kind of detection optical fundus figure
The method of hemorrhagic areas in Xiang.First eye fundus image is split by the method by threshold value based on relative entropy, then uses morphology top
The method of cap conversion extracts blood vessel, finally classifies erythema zone by support vector machine.The method has some limitation: optical fundus figure
In Xiang, it is common scenario that hemorrhagic areas is connected with blood vessel, carries out morphology top cap conversion, put on the error basis of Threshold segmentation
Big this error, produces bad effect.It addition, different because of the state of an illness of patient, in eye fundus image, the feature of lesion region is a lot,
Can be divided into many types, support vector machine is made by quadratic programming and solves support vector, for large-scale training sample, expends
Certain amount of calculation, its arithmetic speed aspect there is also bigger room for promotion.
In addition, deliver in 2009 at Saiprasad Ravishankar, Arpit Jain, Anurag Mittal et al.
Automated Feature Extraction for Early Detection of Diabetic Retinopathy in Fundus Images (eye
Diabetes detection Automatic signature extraction in early days in base map picture) in, it is provided that the method for a series of detection eye fundus image pathological changes, ooze rigid
In the automatically detection gone out, article first use twice expansive working and with both do to differ from and find seepage areas border, then pass through morphology
Fill and obtain candidate region, classify finally by seepage areas gray feature.The method is limited in that, border may have disconnected
Split and cause padding difficulty, and then omit some lesion region.
Based on above technology, need badly and find a kind of new oozing out and hemorrhage automatic testing method and Computational frame, to promote
The optimum effect being accurately positioned out lesion region is reached while arithmetic speed and robustness.
Summary of the invention
The present invention is directed to prior art some weak skirt response inaccuracy, the hemorrhagic areas detection to being connected with blood vessel are the most very managed
Think, and computationally intensive that the deficiency that the aspect such as arithmetic speed exists proposes the inspection of a kind of eye fundus image pathological changes based on morphological segment
Survey method, utilizes the gradient information of image, half-tone information and the feature of blood vessel similar gray value, by oozing out detection algorithm and hemorrhage
Detection algorithm, carries out feature extraction with morphology means, eliminates the complicated methods such as classification, had both improve computing real-time, again
Exactness without misalignment, adapting to while variety classes eye fundus image, it is to avoid the blood vessel of similar gray value is in other hemorrhage detection algorithm
The interference that may bring, accelerates detection speed the most simultaneously.
The present invention is achieved by the following technical solutions:
The present invention relates to a kind of eye fundus image lesion detection method based on morphological segment, first eye fundus image is carried out smooth filter
Ripple, and use region-growing method to position optic disc region；Then obtain removing the optical fundus background of seepage areas by Morphological scale-space
Image, and obtain including blood vessel and hemorrhage Mixed Zone image by Threshold segmentation；Edge inspection is carried out finally by Kirsch operator
Surveying and obtain blood-vessel image by region growing, it is hemorrhagic areas image with the difference of Mixed Zone image and is calculated this district
Territory area.
Described eye fundus image is the most preprocessed, and this pretreatment includes HSV color space brightness correction and based on limiting contrast
The contrast of degree histogram equalization strengthens.
Described Morphological scale-space includes: expansive working is removed blood vessel, found border, morphology filling covering by gradient operator
Seepage areas, iteration etching operation obtain background etc..
Technique effect
Compared with prior art, the material that patient is harmful to by the present invention without using fluorescent agent etc. just can relatively accurately zoning
Area；The eye fundus image being not quite similar for the brightness of different instruments shootings, contrast can be accomplished preferably to adapt to, it is only necessary to adjusts
One or two parameter can accomplish accurately to detect；Compared with the fixed threshold in existing method, partial threshold is taked and image maximum gray scale
Form of ratios be given, universality is higher, is not required to change the size of threshold value for major part image (different brightness, gray scale etc.).
Accompanying drawing explanation
Fig. 1 is present configuration schematic diagram.
Fig. 2 is schematic flow sheet of the present invention.
Fig. 3 is Kirsch operator edge extracting template schematic diagram in embodiment.
Fig. 4 is eye fundus image lesion detecting system MATLAB interface schematic diagram in embodiment.
Fig. 5 be in embodiment morphology fill after result schematic diagram.
Fig. 6 is the background picture schematic diagram generated after corrosion reconstruction iteration in embodiment.
Fig. 7 is hard exudate testing result schematic diagram in embodiment.
Fig. 8 is hemorrhage testing result schematic diagram in embodiment.
Detailed description of the invention
Elaborating embodiments of the invention below, the present embodiment is implemented under premised on technical solution of the present invention,
Give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.
Embodiment 1
As it is shown in figure 1, the present embodiment comprises the following steps:
The first step, input eye fundus image to be detected, and carry out pretreatment, particularly as follows:
1.1) HSV space luminance component is utilized to realize luminance proportion: first eye fundus image to be converted into HSV space, by therein
V component proceeds as follows:Wherein: XVRepresent the V component of pixel, X 'VRepresent after updating
The V component value of pixel；Then by X 'VReturn the rgb space of this pixel, i.e. complete luminance proportion.
1.2) realize contrast by restriction contrast self-adapting histogram equilibrium (CLAHE) to strengthen；
This operation both met self-adapting histogram equilibrium by local histogram redistribution brightness thus reach contrast increase
Strong effect, overcomes again the problem excessively amplifying noise of common self-adapting histogram equilibrium, at eye fundus image lesion region area
Time less, the suppression of noise is a need for.
Optic disc region is positioned on second step, image after the pre-treatment, particularly as follows:
2.1) operator of 31*31 is used to carry out mean filter, the location of mistake that elimination background and lesion region may cause.
2.2) higher than background value due to optic disc gray value, grey level histogram after the filtering is chosen maximum as optic disc anchor point.
2.3) utilize region-growing method, start progressively to expand until optic disc border from anchor point, it is achieved all optic disc regions are labeled
Going out, this region-growing method refers to:
3rd step, for .... carry out hard exudate region detection, particularly as follows:
3.1) by morphology closed operation, the impact that detection is produced by blood vessel is eliminated, particularly as follows:
3.2) by the operator of 11*11, the neighborhood of pixel each in image is carried out variance calculating, choose variance more than boundary threshold
Point as seepage areas boundary candidates point.
3.3) utilize above-mentioned border to carry out morphology filling, cover all seepage areas；Carry out morphological dilations behaviour the most further
Make, obtain image as shown in Figure 5.
Described morphology is filled and is realized by the imfill () function in MATLAB.
3.4) carry out morphological erosion repeatedly and reconstruction operation, background gray scale is filled in seepage areas step by step, obtains
The background image without pathological changes as shown in Figure 6.
3.5) with original image, the optical fundus background image removing seepage areas being done difference, the bigger part of difference is seepage areas, will
Matrix of differences carries out Threshold segmentation, and is marked in artwork by qualified point, zoning area simultaneously.
4th step, carry out hemorrhagic areas detection, particularly as follows:
4.1) according to blood vessel and hemorrhagic areas gray threshold, pretreated image is carried out Threshold segmentation, obtain including blood vessel and
Hemorrhage Mixed Zone image.
4.2) by Kirsch operator, the vessel boundary in the image of Mixed Zone is marked, particularly as follows:
4.2.1) as it is shown on figure 3, arrange four 3x3 templates, four templates press 0,45,90,135 respectively so that (x y) is
The region of 3x3 is divided into two parts by center, and each pixel in image carries out respectively convolution summation behaviour according to these four templates
Make.
4.2.2) four results seeking pixel each in image seek absolute value, by each result respectively with a threshold ratio relatively,
If the most any one result is more than or equal to threshold value T, then the gray value of the image slices vegetarian refreshments corresponding to the central point of this template is
255, it is otherwise 0.
4.3) blood-vessel image in the image of Mixed Zone is extracted by region-growing method, particularly as follows:
4.4) blood-vessel image extracted is rejected from the image of Mixed Zone, then eliminates tiny noise by morphology closed operation,
Thus obtain hemorrhagic areas image.
5th step, hemorrhagic areas image is gone out with different color markings on eye fundus image to be detected, and pass through lesion region
The ratio of pixel quantity and the pixel quantity of whole effective image-region is worth to the area of corresponding region.
According to above-mentioned steps, we are tested on a DELL microcomputer, and the major parameter of this computer is: centre
Reason device Intel (R) Core (TM) i5 4210U, 8.00GB internal memory, AMD Radeon R7M265 series video card, Windows
864 bit manipulation systems.Fig. 7, Fig. 8 describe our experimental result.
Claims (1)
1. an eye fundus image lesion detection method based on morphological segment, it is characterised in that first eye fundus image is smoothed
Filtering, and use region-growing method to position optic disc region；Then obtain removing the optical fundus back of the body of seepage areas by Morphological scale-space
Scape image, and obtain including blood vessel and hemorrhage Mixed Zone image by Threshold segmentation；Edge is carried out finally by Kirsch operator
Detecting and obtain blood-vessel image by region growing, it is hemorrhagic areas image with the difference of Mixed Zone image and is calculated this
Region area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510126237.9A CN106157279A (en) | 2015-03-23 | 2015-03-23 | Eye fundus image lesion detection method based on morphological segment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510126237.9A CN106157279A (en) | 2015-03-23 | 2015-03-23 | Eye fundus image lesion detection method based on morphological segment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106157279A true CN106157279A (en) | 2016-11-23 |
Family
ID=58063073
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510126237.9A Pending CN106157279A (en) | 2015-03-23 | 2015-03-23 | Eye fundus image lesion detection method based on morphological segment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106157279A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106725295A (en) * | 2016-11-29 | 2017-05-31 | 瑞达昇科技(大连)有限公司 | A kind of miniature check-up equipment, device and its application method |
| CN108846827A (en) * | 2018-04-16 | 2018-11-20 | 江南大学 | A method of based on more circle Fast Segmentation eyeground optic disks |
| CN109410191A (en) * | 2018-10-18 | 2019-03-01 | 中南大学 | Optical fundus blood vessel localization method and its anaemia screening method based on OCT image |
| CN109816637A (en) * | 2019-01-02 | 2019-05-28 | 电子科技大学 | The detection method in hard exudate region in a kind of eye fundus image |
| CN110363739A (en) * | 2018-04-08 | 2019-10-22 | 天津工业大学 | Eye fundus image hard exudate detection method based on background estimating and phase equalization |
| WO2019218118A1 (en) * | 2018-05-14 | 2019-11-21 | 深圳明眸科技有限公司 | Fundus oculi lesion area calculation method, apparatus, medical device, and storage medium |
| CN111311565A (en) * | 2020-02-11 | 2020-06-19 | 平安科技（深圳）有限公司 | Eye OCT image-based detection method and device for positioning points of optic cups and optic discs |
| CN112106146A (en) * | 2018-03-08 | 2020-12-18 | 皇家飞利浦有限公司 | Interactive self-improving annotation system for high-risk plaque burden assessment |
| CN112699841A (en) * | 2021-01-13 | 2021-04-23 | 华南理工大学 | Traffic sign detection and identification method based on driving video |
| CN112990367A (en) * | 2021-04-25 | 2021-06-18 | 杭州晟视科技有限公司 | Image processing method, device, equipment and storage medium |
| CN116309549A (en) * | 2023-05-11 | 2023-06-23 | 爱尔眼科医院集团股份有限公司 | Fundus region detection method, fundus region detection device, fundus region detection equipment and readable storage medium |
-
2015
- 2015-03-23 CN CN201510126237.9A patent/CN106157279A/en active Pending
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106725295A (en) * | 2016-11-29 | 2017-05-31 | 瑞达昇科技(大连)有限公司 | A kind of miniature check-up equipment, device and its application method |
| CN112106146A (en) * | 2018-03-08 | 2020-12-18 | 皇家飞利浦有限公司 | Interactive self-improving annotation system for high-risk plaque burden assessment |
| CN110363739A (en) * | 2018-04-08 | 2019-10-22 | 天津工业大学 | Eye fundus image hard exudate detection method based on background estimating and phase equalization |
| CN108846827B (en) * | 2018-04-16 | 2021-10-15 | 江南大学 | Method for rapidly segmenting fundus optic disk based on multiple circles |
| CN108846827A (en) * | 2018-04-16 | 2018-11-20 | 江南大学 | A method of based on more circle Fast Segmentation eyeground optic disks |
| WO2019218118A1 (en) * | 2018-05-14 | 2019-11-21 | 深圳明眸科技有限公司 | Fundus oculi lesion area calculation method, apparatus, medical device, and storage medium |
| CN109410191A (en) * | 2018-10-18 | 2019-03-01 | 中南大学 | Optical fundus blood vessel localization method and its anaemia screening method based on OCT image |
| CN109410191B (en) * | 2018-10-18 | 2022-03-25 | 中南大学 | OCT (optical coherence tomography) image-based fundus blood vessel positioning method and anemia screening method thereof |
| CN109816637A (en) * | 2019-01-02 | 2019-05-28 | 电子科技大学 | The detection method in hard exudate region in a kind of eye fundus image |
| CN111311565A (en) * | 2020-02-11 | 2020-06-19 | 平安科技（深圳）有限公司 | Eye OCT image-based detection method and device for positioning points of optic cups and optic discs |
| CN112699841A (en) * | 2021-01-13 | 2021-04-23 | 华南理工大学 | Traffic sign detection and identification method based on driving video |
| CN112990367A (en) * | 2021-04-25 | 2021-06-18 | 杭州晟视科技有限公司 | Image processing method, device, equipment and storage medium |
| CN116309549A (en) * | 2023-05-11 | 2023-06-23 | 爱尔眼科医院集团股份有限公司 | Fundus region detection method, fundus region detection device, fundus region detection equipment and readable storage medium |
| CN116309549B (en) * | 2023-05-11 | 2023-10-03 | 爱尔眼科医院集团股份有限公司 | Fundus region detection method, fundus region detection device, fundus region detection equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106157279A (en) | Eye fundus image lesion detection method based on morphological segment | |
| CN110458883B (en) | Medical image processing system, method, device and equipment | |
| CN110276356B (en) | Fundus image microaneurysm identification method based on R-CNN | |
| CN111931811B (en) | Calculation method based on super-pixel image similarity | |
| Zhang et al. | Automated semantic segmentation of red blood cells for sickle cell disease | |
| CN107180421B (en) | Fundus image lesion detection method and device | |
| NL2024774B1 (en) | Blood leukocyte segmentation method based on adaptive histogram thresholding and contour detection | |
| Harangi et al. | Automatic exudate detection using active contour model and regionwise classification | |
| CN105825516A (en) | DICOM image blood flow analysis system | |
| CN105261015B (en) | Eye fundus image blood vessel automatic division method based on Gabor filter | |
| KR102250688B1 (en) | Method and device for automatic vessel extraction of fundus photography using registration of fluorescein angiography | |
| CN110729045A (en) | Tongue image segmentation method based on context-aware residual error network | |
| CN108257124A (en) | A kind of white blood cell count(WBC) method and system based on image | |
| CN104036521A (en) | Novel retina eye fundus image segmenting method | |
| CN111326238A (en) | Cancer cell detection device based on sliding window | |
| CN106384343A (en) | Morphological processing based hard exudation area detecting method | |
| Kaldera et al. | MRI based glioma segmentation using deep learning algorithms | |
| KR20200093502A (en) | Method and device for automatic vessel extraction of fundus photography using registration of fluorescein angiography | |
| Ali et al. | Optic Disc Localization in Retinal Fundus Images Based on You Only Look Once Network (YOLO). | |
| CN109816665A (en) | A kind of fast partition method and device of optical coherence tomographic image | |
| CN105225234A (en) | Based on the lung tumor identification method of support vector machine MRI Iamge Segmentation | |
| Nugroho et al. | Multithresholding approach for segmenting plasmodium parasites | |
| CN116486398A (en) | Focal image extraction method and device, electronic equipment and storage medium | |
| Yang et al. | Detection of microaneurysms and hemorrhages based on improved Hessian matrix | |
| CN105184799A (en) | Modified non-supervision brain tumour MRI (Magnetic Resonance Imaging) image segmentation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20161123 |