CN109615614A - The extracting method and electronic equipment of eye fundus image medium vessels based on multi-feature fusion - Google Patents
The extracting method and electronic equipment of eye fundus image medium vessels based on multi-feature fusion Download PDFInfo
- Publication number
- CN109615614A CN109615614A CN201811418985.4A CN201811418985A CN109615614A CN 109615614 A CN109615614 A CN 109615614A CN 201811418985 A CN201811418985 A CN 201811418985A CN 109615614 A CN109615614 A CN 109615614A
- Authority
- CN
- China
- Prior art keywords
- image
- blood vessel
- feature
- optical fundus
- eye fundus
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-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
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- 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
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Eye Examination Apparatus (AREA)
Abstract
The embodiment of the present invention provides the extracting method and electronic equipment of a kind of eye fundus image medium vessels based on multi-feature fusion, the method comprise the steps that being based on eye fundus image to be processed, utilize multiple and different classifiers, multiple and different features of optical fundus blood vessel are extracted respectively, and Fusion Features are carried out to the multiple different characteristic；Based on the comprehensive characteristics after Fusion Features, the denseCRF model completed using training obtains the segmented image of the optical fundus blood vessel；Morphological analysis processing is carried out to the segmented image, extracts the bianry image of the optical fundus blood vessel.The embodiment of the present invention extracts the blood-vessel image in eye fundus image, can more effectively, more accurately extract vessel boundary image, the optical fundus blood vessel edge image including blurred picture by the multiple features fusion to optical fundus blood vessel.
Description
Technical field
The present embodiments relate to technical field of image processing, more particularly, to a kind of eye based on multi-feature fusion
Extracting method and electronic equipment of the base map as medium vessels.
Background technique
The method of analysis for eye fundus image, image, semantic segmentation is more and more of interest by researcher, such as threshold
It is worth split plot design, edge split plot design, gene code division method, wavelet transformation split plot design, cluster segmentation method etc..As artificial intelligence is known
The development of other technology, neural network (CNN) have caused researcher and have widely paid close attention to, and are applied to image segmentation field.
Currently, more is to be with FCN+denseCRF based on the image, semantic dividing method of full convolutional network FCN+CRF
The image segmentation algorithm of representative, training sample obtain the optical fundus blood vessel space shape comprising non-pointed boundary by training FCN model
Shape divides feature, carries out denseCRF training with this segmentation feature, obtains optical fundus blood vessel segmentation figure.Due to trained aspect ratio
It is more single, so that the classification results that the model trained obtains are not fine enough, it is insensitive to the details in image, so extract
Precision is not good enough, not high for edge definition, insufficient especially for the vessel boundary image zooming-out details of some blurred pictures.
Summary of the invention
In order to overcome the above problem or at least be partially solved the above problem, the embodiment of the present invention provides a kind of based on more
The extracting method and electronic equipment of the eye fundus image medium vessels of Fusion Features, to it is more effective, more accurately extract vessel boundary
Image, the optical fundus blood vessel edge image including blurred picture.
In a first aspect, the embodiment of the present invention provides a kind of extraction side of eye fundus image medium vessels based on multi-feature fusion
Method, comprising:
Based on eye fundus image to be processed, using multiple and different classifiers, multiple and different features of optical fundus blood vessel are extracted respectively,
And Fusion Features are carried out to the multiple different characteristic；
Based on the comprehensive characteristics after Fusion Features, the denseCRF model completed using training obtains the optical fundus blood vessel
Segmented image；
Morphological analysis processing is carried out to the segmented image, extracts the bianry image of the optical fundus blood vessel.
Second aspect, the embodiment of the present invention provide a kind of extraction dress of eye fundus image medium vessels based on multi-feature fusion
It sets, comprising:
Comprehensive characteristics extraction module, for extracting eye respectively using multiple and different classifiers based on eye fundus image to be processed
Multiple and different features of bottom blood vessel, and Fusion Features are carried out to the multiple different characteristic；
Optical fundus blood vessel image segmentation module, for being completed using training based on the comprehensive characteristics after Fusion Features
DenseCRF model obtains the segmented image of the optical fundus blood vessel；
Output module is extracted, for carrying out morphological analysis processing to the segmented image, extracts the optical fundus blood vessel
Bianry image.
The third aspect, the embodiment of the present invention provide a kind of electronic equipment, comprising: at least one processor, at least one
Manage device, communication interface and bus；The memory, the processor and the communication interface are completed mutual by the bus
Communication, the communication interface between the electronic equipment and eye fundus image equipment information transmission；In the memory
It is stored with the computer program that can be run on the processor, when the processor executes the computer program, is realized such as
The extracting method of eye fundus image medium vessels based on multi-feature fusion described in upper first aspect.
Fourth aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, the non-transient calculating
Machine readable storage medium storing program for executing stores computer instruction, and the computer instruction executes the computer described in first aspect as above
The extracting method of eye fundus image medium vessels based on multi-feature fusion.
The extracting method and electronic equipment of eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention,
The complementarity between recognition result that consideration different classifications device obtains, Various Classifiers on Regional is blended, and various features are merged
It is combined with FCN+denseCRF model, carries out the extraction of eye fundus image medium vessels image, the segmentation of optical fundus blood vessel can be made to mention
Take effect more preferable, the image of fuzzy edge can also extract part blood vessel, improve extraction accuracy.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the stream of the extracting method for the eye fundus image medium vessels based on multi-feature fusion that one embodiment of the invention provides
Journey schematic diagram；
Fig. 2 is in the extracting method according to eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention
Carry out the flow diagram of multiple features fusion；
Fig. 3 is in the extracting method according to eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention
The flow diagram of morphological analysis is carried out to segmented image；
Fig. 4 be another embodiment of the present invention provides eye fundus image medium vessels based on multi-feature fusion extracting method
Flow diagram；
Fig. 5 is the structure of the extraction element of eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention
Schematic diagram；
Fig. 6 is the entity structure schematic diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the embodiment of the present invention, instead of all the embodiments.Based on the embodiment in the embodiment of the present invention, ability
Domain those of ordinary skill every other embodiment obtained without making creative work, belongs to the present invention
The range of embodiment protection.
Since neural network has very strong adaptivity and learning ability, non-linear mapping capability, robustness and fault-tolerant
Ability receives people and more and more payes attention to.However, for using high dimensional feature to carry out image recognition, traditional individually divides
Class device is difficult to obtain satisfactory discrimination.It is often complementary very strong between the recognition result obtained due to different classifications device,
Various Classifiers on Regional is blended, the network model of training can increase the available quantity of identification information after various features merge,
The uncertainty for reducing information is to improve whole system discrimination, accuracy effective way.The embodiment of the present invention uses a variety of spies
The mode that sign fusion and FCN+denseCRF model combine, keeps the segmentation extraction effect of optical fundus blood vessel more preferable, fuzzy edge
Image can also extract part blood vessel, improve precision.Below will especially by multiple embodiments to the embodiment of the present invention into
Row expansion explanation and introduction.
Fig. 1 is the stream of the extracting method for the eye fundus image medium vessels based on multi-feature fusion that one embodiment of the invention provides
Journey schematic diagram, as shown in Figure 1, this method comprises:
S101 is based on eye fundus image to be processed, using multiple and different classifiers, extracts the multiple and different of optical fundus blood vessel respectively
Feature, and Fusion Features are carried out to multiple different characteristic.
The complementarity between recognition result that the embodiment of the present invention is obtained using different classifications device, Various Classifiers on Regional is mutually melted
It closes.That is, different classifiers is respectively adopted, the corresponding different characteristic for extracting optical fundus blood vessel from eye fundus image to be processed, these
Different characteristic characterizes the shape of optical fundus blood vessel from different perspectives.Wherein optional, above-mentioned multiple and different features specifically can be
Edge feature, textural characteristics and the spatial form feature of optical fundus blood vessel.
Later, for these different characteristics, to keep the standard to optical fundus blood vessel image more acurrate, by these different characteristics into
Row Fusion Features, the comprehensive characteristics after obtaining Fusion Features.It is understood that can be adopted when carrying out different characteristic fusion
It is carried out, can also be merged using some innovatory algorithms with existing some data anastomosing algorithms.
S102, based on the comprehensive characteristics after Fusion Features, the denseCRF model completed using training obtains optical fundus blood vessel
Segmented image.
The embodiment of the present invention needs after the comprehensive characteristics for obtaining optical fundus blood vessel image according to above-mentioned processing according to the synthesis
Feature obtains the segmented image of optical fundus blood vessel from eye fundus image to be processed.Specifically, can use condition random field
(denseCRF) divide to realize.Using each pixel as node, the relationship between pixel and pixel constitutes one as side
A condition random field.Using the comprehensive characteristics after variant Fusion Features, global classification transfer is considered in conjunction with denseCRF structure
Influence, realize the segmentation of optical fundus blood vessel image, can be improved accuracy, optimum results optimized after optical fundus blood vessel segmentation figure
Picture.
S103 carries out morphological analysis processing to segmented image, extracts the bianry image of optical fundus blood vessel.
The embodiment of the present invention is on the basis of obtaining the segmented image of optical fundus blood vessel according to above-mentioned processing, for intuitive characterization
The shape of optical fundus blood vessel and position etc. carry out morphological analysis processing according to the segmented image of optical fundus blood vessel, finally obtain wait divide
The bianry image of the optical fundus blood vessel cut.
It is understood that wherein morphological analysis be mainly used for from image extract to expression and description region shape have
The picture content of meaning enables subsequent identification work to catch the shape feature of target object most separating capacity, such as boundary
With connected region etc..
The extracting method of eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention considers different points
The complementarity between recognition result that class device obtains, Various Classifiers on Regional is blended, and by various features fusion and FCN+
DenseCRF model combines, and carries out the extraction of eye fundus image medium vessels image, can make the segmentation extraction effect of optical fundus blood vessel
More preferably, the image of fuzzy edge can also extract part blood vessel, improve extraction accuracy.
Wherein, according to the above embodiments optionally, edge feature, textural characteristics and the space of optical fundus blood vessel are extracted respectively
The step of shape feature, specifically includes: utilizing Canny algorithm classification device, analyzes eye fundus image to be processed, extracts edge feature；Benefit
With LBP algorithm classification device, eye fundus image to be processed, texture feature extraction are analyzed；And using FCN algorithm classification device, analyze wait locate
Eye fundus image is managed, spatial form feature is extracted.
It is understood that in view of the property of the different feature of the various embodiments described above optical fundus blood vessel, i.e. optical fundus blood vessel
The heterogeneity of edge feature, textural characteristics and spatial form feature selects corresponding classifier from eye fundus image to be processed
Carry out the extraction respectively of these features.It extracts the textural characteristics of optical fundus blood vessel specifically, being based respectively on LBP, mentioned based on canny
The optical fundus blood vessel spatial form segmentation comprising non-pointed boundary that the optical fundus blood vessel edge feature and FCN taken extracts is special
Sign.The pixel that full convolution layer network FCN model therein is directly based upon image is operated, by a series of convolutional layer and
One layer of warp lamination exports class probability finally by softmax layers and obtains the segmented image of segmentation feature.
It is understood that in order to realize accurate optical fundus blood vessel image segmentation, need using above-mentioned each classifier i.e.
Handle model carry out data processing before, using a certain amount of training sample to these classifiers i.e. handle model be trained and
Optimization updates, i.e., before the step of utilizing Canny algorithm classification device, analyzing eye fundus image to be processed, extract edge feature, this
The method of inventive embodiments further include: a certain number of eye fundus image training samples are obtained, and utilize eye fundus image training sample,
To the basic Canny algorithm classification device of foundation, basis LBP algorithm classification device, basis FCN algorithm classification device and basis denseCRF
Model is iterated training, obtains Canny algorithm classification device, LBP algorithm classification device, FCN algorithm classification device and denseCRF mould
Type.
It is understood that basis Canny algorithm classification device therein, basis LBP algorithm classification device, basis FCN algorithm
Classifier and basis denseCRF model can be in advance according to application demand Stochastic Models parameter or initial setting up mould
After shape parameter, the corresponding network model that automatically generates.Later, need using a certain amount of training pattern to these network models into
Row iteration training is updated with optimization, the optimal model parameters to impose a condition until obtaining satisfaction, namely obtains optimal models
It is using the optimal models as the final processing model for carrying out optical fundus blood vessel image segmentation, i.e., applied according to the above embodiments
Canny algorithm classification device, LBP algorithm classification device, FCN algorithm classification device and denseCRF model.
Wherein, according to the above embodiments optionally, the step of carrying out Fusion Features to multiple and different features specifically includes:
It is based respectively on edge feature, textural characteristics and spatial form feature, extracts feature vector, and each feature vector of extraction is carried out
Cascade constitutes multi-feature vector；Using probabilistic standard algorithm, multi-feature vector is normalized, and is utilized
Principal component analysis algorithm handles the multi-feature vector after normalization, realizes Fusion Features.
Fig. 2 is in the extracting method according to eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention
The flow diagram of multiple features fusion is carried out, as shown in Fig. 2, the embodiment of the present invention is carrying out the multiple features of optical fundus blood vessel image
When fusion, it is primarily based on edge feature, textural characteristics and the spatial form feature of the optical fundus blood vessel of different classifications device extraction, is extracted
The feature vector of these different characteristics, and these feature vectors are cascaded, form multi-feature vector X=
(X1,···,Xn).Later, using probabilistic standard algorithm, to multi-feature vector X=(X1,···,Xn) carry out normalizing
Change processing, multi-feature vector X '=(X ' after being normalized1,...X′n), so that the order of magnitude of each component is close to one
It causes.Finally, using Principal Component Analysis Method, by solving the autocorrelation matrix of the multi-feature vector after normalizing and carrying out corresponding
Karhunen-Loeve transformation, solution obtain the Fusion Features of above-mentioned variant feature.
Fig. 3 is in the extracting method according to eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention
The flow diagram that morphological analysis is carried out to segmented image, as shown in figure 3, using principal component analysis algorithm, after normalization
Multi-feature vector handled, realize Fusion Features the step of specifically include: seek normalization after multi-feature vector
Autocorrelation matrix, and autocorrelation matrix is decomposed using principal component analysis algorithm, obtain respectively eigenvectors matrix and
Eigenvalue matrix；Multi-feature vector after normalization is subjected to Karhunen-Loeve transformation, obtains transformed eigenvalue matrix, and choose
Feature vector corresponding to maximum eigenvalue in transformed eigenvalue matrix, as comprehensive characteristics.
Eyeground is realized using Principal Component Analysis Method based on the multi-feature vector after normalization according to above-described embodiment
Multi-feature vector X '=(X ' when the fusion of blood vessel different characteristic, after seeking normalization first1,...X′n) auto-correlation square
Battle array R carries out feature decomposition to R, obtains eigenvectors matrix A and characteristic value square further according to the principle of Principal Component Analysis Method (PCA)
Battle array U.Then, to X '=(X '1,...X′n) Karhunen-Loeve transformation is carried out, Y '=U ' X ' is such as enabled, then Y ' is to X '=(X '1,...X′n)
Karhunen-Loeve transformation.Wherein each column vector of U ' contains multi-feature vector X=(X1,···,Xn) information including, it
It afterwards can fusion feature by Y ' as X.Furthermore it is possible to choose the corresponding feature vector of maximum eigenvalue as the comprehensive special of X
Sign, achievees the purpose that Fusion Features, obtains the comprehensive characteristics comprising optical fundus blood vessel texture, edge and spatial form.
Wherein, according to the above embodiments optionally, morphological analysis processing is carried out to segmented image, extracts optical fundus blood vessel
Bianry image the step of specifically include: to segmented image carry out gray processing negate, and to the negated image of gray processing carry out
The high cap filtering processing of form, the image after obtaining cap transformation；Using Threshold Analysis method, two are carried out to the image after cap transformation
Value processing, obtains the bianry image of optical fundus blood vessel.
The embodiment of the present invention carries out gray processing processing to segmented image according to the above embodiments first and negates,
Obtain the negated image of gray processing.Through this, treated that vasculature part gray scale is larger (brighter), and then gray scale is smaller for background parts
(darker) facilitates next high cap filtering processing.
Later, the high cap of form is carried out to the negated image of the gray processing of input to filter, i.e., morphology is subtracted from image
Image after opening operation, the image after obtaining cap transformation.Picture contrast can be enhanced by the filtering of high cap.To bianry image
Or gray level image can take high cap to filter.Picture contrast can be enhanced by the filtering of high cap, cap transformation can be by bright mesh
Mark reveals to come from dark background convexity.
Finally, carrying out thresholding to the image after cap transformation.That is, being carried on the back using the target area to be extracted in image with it
Difference of the scape in gamma characteristic regards image as in the two class regions (target area and background area) with different grey-scale
Combination, chooses the reasonable threshold value of comparison, each pixel should belong to target area or background area in image to determine
Domain, to generate corresponding bianry image.By the way that threshold parameter is arranged, to extract the bianry image of optical fundus blood vessel.
For the technical solution that the embodiment of the present invention will be further explained, the embodiment of the present invention provides such as according to the above embodiments
The process flow of lower embodiment, but the protection scope of the embodiment of the present invention is not limited.
Fig. 4 be another embodiment of the present invention provides eye fundus image medium vessels based on multi-feature fusion extracting method
Flow diagram, as shown in figure 4, the extracting method of the embodiment of the present invention includes following treatment process:
Firstly, using eye fundus image training sample to FCN algorithm classification device, LBP algorithm classification device and Canny algorithm classification
Device is trained, and the available optical fundus blood vessel spatial form comprising non-pointed boundary divides feature, optical fundus blood vessel textural characteristics
With optical fundus blood vessel edge feature, these optical fundus blood vessel features are subjected to multiple features fusion later.That is,
Be based respectively on LBP extract optical fundus blood vessel textural characteristics, based on canny extract optical fundus blood vessel edge feature and
The optical fundus blood vessel spatial form segmentation feature comprising non-pointed boundary that FCN is extracted；
The feature vector of three kinds of features is extracted again, these features are then formed into vector X=(X1···Xn)；
Probabilistic standard method is recycled, to feature vector, X=(X1···Xn) be normalized, so that each component
The order of magnitude close to consistent；
Finally seek feature vector, X '=(X ' after normalization1,...X′n) autocorrelation matrix R, according to the original of PCA
Reason carries out feature decomposition to R, obtains eigenvectors matrix A and eigenvalue matrix U, and X ' is then passed through Karhunen-Loeve transformation, transformed
Including each column vector of U ' in journey contains the information of X, later can fusion feature by the Y ' of Karhunen-Loeve transformation as X,
Comprehensive characteristics of the corresponding feature vector of maximum eigenvalue as X can be chosen, achieve the purpose that Fusion Features, are obtained comprising eye
Bottom vascular lake, edge, spatial form comprehensive characteristics.
Secondly, with obtained comprehensive characteristics training denseCRF model, the segmented image of available optical fundus blood vessel.It can be with
Understand, the above-mentioned various features fusion for eye fundus image blood vessel segmentation can also be used among other network models.
Morphological scale-space operation is carried out finally, for the optical fundus blood vessel segmented image obtained by denseCRF model, most
Optical fundus blood vessel bianry image is obtained eventually.
The embodiment of the present invention is for extracting the textural characteristics of optical fundus blood vessel, the optical fundus blood vessel extracted based on canny based on LBP
The optical fundus blood vessel spatial form segmentation feature comprising non-pointed boundary that edge feature and FCN are extracted carries out PCA processing
Various features are merged, go to train denseCRF mould with the obtained comprehensive characteristics comprising optical fundus blood vessel texture, edge, spatial form
Type.It is often complementary very strong between the recognition result obtained due to different classifications device, Various Classifiers on Regional is blended, process is a variety of
The network model of training can increase the available quantity of identification information after Fusion Features, reduce the uncertainty of information, be that raising is whole
A system recognition rate, accuracy effective way.Present invention employs various features fusions and FCN+denseCRF model to combine
Keep the segmentation extraction effect of optical fundus blood vessel more preferable, fuzzy pathological image can also extract part vessel boundary, improve blood
The extraction accuracy of pipe and vessel boundary.
As the other side of the embodiment of the present invention, the embodiment of the present invention provides one kind according to the above embodiments and is based on
The extraction element of the eye fundus image medium vessels of multiple features fusion, the device are based on multiple features for realizing in the above embodiments
The extraction of the eye fundus image medium vessels of fusion.Therefore, the blood in the eye fundus image based on multi-feature fusion of the various embodiments described above
Description and definition in the extracting method of pipe, can be used for the understanding of each execution module in the embodiment of the present invention, can specifically join
Above-described embodiment is examined, is not being repeated herein.
One embodiment according to an embodiment of the present invention, the extraction element of eye fundus image medium vessels based on multi-feature fusion
Structure as shown in figure 5, be eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention extraction element
Structural schematic diagram, which can be used to implement in above-mentioned each method embodiment blood in eye fundus image based on multi-feature fusion
The extraction of pipe, the device include: comprehensive characteristics extraction module 501, optical fundus blood vessel image segmentation module 502 and extraction output module
503.Wherein:
Comprehensive characteristics extraction module 501 is used to extract respectively based on eye fundus image to be processed using multiple and different classifiers
Multiple and different features of optical fundus blood vessel, and Fusion Features are carried out to multiple and different features；Optical fundus blood vessel image segmentation module 502 is used
In based on the comprehensive characteristics after Fusion Features, the denseCRF model completed using training obtains the segmented image of optical fundus blood vessel；
It extracts output module 503 to be used to carry out morphological analysis processing to segmented image, extracts the bianry image of optical fundus blood vessel.
Specifically, different classifiers is respectively adopted in comprehensive characteristics extraction module 501, it is right from eye fundus image to be processed
The different characteristic of optical fundus blood vessel should be extracted, these different characteristics characterize the shape of optical fundus blood vessel from different perspectives.It is wherein optional,
Above-mentioned multiple and different features specifically can be the edge feature, textural characteristics and spatial form feature of optical fundus blood vessel.Later, right
In these different characteristics, comprehensive characteristics extraction module 501 is to keep the standard to optical fundus blood vessel image more acurrate, these are different special
Sign carries out Fusion Features, the comprehensive characteristics after obtaining Fusion Features.
Optical fundus blood vessel image segmentation module 502 can use condition random field (denseCRF) to realize segmentation.With each
For pixel as node, the relationship between pixel and pixel constitutes a condition random field as side.Optical fundus blood vessel image point
Module 502 is cut using the comprehensive characteristics after variant Fusion Features, considers that global classification transfer influences in conjunction with denseCRF structure,
Realize the segmentation of optical fundus blood vessel image.
For shape and the position etc. for intuitively characterizing optical fundus blood vessel, output module 503 is extracted according to the segmentation of optical fundus blood vessel
Image carries out morphological analysis processing, finally obtains the bianry image of optical fundus blood vessel to be split.
The extraction element of eye fundus image medium vessels based on multi-feature fusion provided in an embodiment of the present invention, by the way that phase is arranged
The execution module answered, the complementarity between recognition result that consideration different classifications device obtains, Various Classifiers on Regional is blended, and will
Various features fusion is combined with FCN+denseCRF model, is carried out the extraction of eye fundus image medium vessels image, can be made eyeground
The segmentation extraction effect of blood vessel is more preferable, and the image of fuzzy edge can also extract part blood vessel, improves extraction accuracy.
It is understood that can be by hardware processor (hardware processor) come real in the embodiment of the present invention
Each relative program module in the device of existing the various embodiments described above.Also, the eye based on multi-feature fusion of the embodiment of the present invention
Base map as medium vessels extraction element utilize above-mentioned each program module, can be realized above-mentioned each method embodiment based on multiple features
The extraction process of the eye fundus image medium vessels of fusion, the eye based on multi-feature fusion in for realizing above-mentioned each method embodiment
Base map as medium vessels extraction when, beneficial effect that the device of the embodiment of the present invention generates and corresponding above-mentioned each method embodiment
It is identical, above-mentioned each method embodiment can be referred to, details are not described herein again.
As the another aspect of the embodiment of the present invention, the present embodiment provides a kind of electronics according to the above embodiments and sets
It is standby, it is the entity structure schematic diagram of electronic equipment provided in an embodiment of the present invention, comprising: at least one processor with reference to Fig. 6
601, at least one processor 602, communication interface 603 and bus 604.
Wherein, memory 601, processor 602 and communication interface 603 complete mutual communication by bus 604, communicate
Interface 603 is for the information transmission between the electronic equipment and eye fundus image equipment；Being stored in memory 601 can be in processor
The computer program run on 602 when processor 602 executes the computer program, realizes the base as described in the various embodiments described above
In the extracting method of the eye fundus image medium vessels of multiple features fusion.
It is to be understood that including at least memory 601, processor 602, communication interface 603 and bus in the electronic equipment
604, and memory 601, processor 602 and communication interface 603 form mutual communication connection by bus 604, and can be complete
Mentioning for eye fundus image medium vessels based on multi-feature fusion is read from memory 601 at mutual communication, such as processor 602
Take the program instruction etc. of method.In addition, communication interface 603 can also realize leading between the electronic equipment and eye fundus image equipment
Letter connection, and achievable mutual information transmission, are such as realized in eye fundus image based on multi-feature fusion by communication interface 603
The extraction etc. of blood vessel.
When electronic equipment is run, processor 602 calls the program instruction in memory 601, real to execute above-mentioned each method
Apply method provided by example, for example, eyeground is extracted using multiple and different classifiers based on eye fundus image to be processed respectively
Multiple and different features of blood vessel, and Fusion Features are carried out to multiple and different features；Based on the comprehensive characteristics after Fusion Features, utilize
The denseCRF model that training is completed, obtains the segmented image of optical fundus blood vessel；Morphological analysis processing is carried out to segmented image, is mentioned
Take the bianry image etc. of optical fundus blood vessel.
Program instruction in above-mentioned memory 601 can be realized and as independent by way of SFU software functional unit
Product when selling or using, can store in a computer readable storage medium.Alternatively, realizing that above-mentioned each method is implemented
This can be accomplished by hardware associated with program instructions for all or part of the steps of example, and program above-mentioned can store to be calculated in one
In machine read/write memory medium, when being executed, execution includes the steps that above-mentioned each method embodiment to the program；And storage above-mentioned
Medium includes: USB flash disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random
Access Memory, RAM), the various media that can store program code such as magnetic or disk.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium also according to the various embodiments described above, this is non-temporarily
State computer-readable recording medium storage computer instruction, the computer instruction execute computer as described in the various embodiments described above
Eye fundus image medium vessels based on multi-feature fusion extracting method, for example, eye fundus image to be processed is based on, using more
A different classifications device extracts multiple and different features of optical fundus blood vessel respectively, and carries out Fusion Features to multiple and different features；It is based on
Comprehensive characteristics after Fusion Features, the denseCRF model completed using training, obtain the segmented image of optical fundus blood vessel；To segmentation
Image carries out morphological analysis processing, extracts the bianry image etc. of optical fundus blood vessel.
Electronic equipment provided in an embodiment of the present invention and non-transient computer readable storage medium, by executing above-mentioned each reality
The extracting method of eye fundus image medium vessels based on multi-feature fusion described in example is applied, considers the identification knot that different classifications device obtains
Complementarity between fruit, Various Classifiers on Regional is blended, and various features are merged and are combined with FCN+denseCRF model, into
The extraction of row eye fundus image medium vessels image, can make the segmentation extraction effect of optical fundus blood vessel more preferable, the image of fuzzy edge
Part blood vessel can be extracted, extraction accuracy is improved.
It is understood that the embodiment of device described above, electronic equipment and storage medium is only schematic
, wherein unit may or may not be physically separated as illustrated by the separation member, it can both be located at one
Place, or may be distributed on heterogeneous networks unit.Some or all of modules can be selected according to actual needs
To achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are without paying creative labor
To understand and implement.
By the description of embodiment of above, those skilled in the art is it will be clearly understood that each embodiment can borrow
Help software that the mode of required general hardware platform is added to realize, naturally it is also possible to pass through hardware.Based on this understanding, above-mentioned
Substantially the part that contributes to existing technology can be embodied in the form of software products technical solution in other words, the meter
Calculation machine software product may be stored in a computer readable storage medium, such as USB flash disk, mobile hard disk, ROM, RAM, magnetic disk or light
Disk etc., including some instructions, with so that a computer equipment (such as personal computer, server or network equipment etc.)
Execute method described in certain parts of above-mentioned each method embodiment or embodiment of the method.
In addition, those skilled in the art are it should be understood that in the application documents of the embodiment of the present invention, term
"include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that including a series of elements
Process, method, article or equipment not only include those elements, but also including other elements that are not explicitly listed, or
Person is to further include for elements inherent to such a process, method, article, or device.In the absence of more restrictions, by
The element that sentence "including a ..." limits, it is not excluded that in the process, method, article or apparatus that includes the element
There is also other identical elements.
In the specification of the embodiment of the present invention, numerous specific details are set forth.It should be understood, however, that the present invention is implemented
The embodiment of example can be practiced without these specific details.In some instances, it is not been shown in detail well known
Methods, structures and technologies, so as not to obscure the understanding of this specification.Similarly, it should be understood that in order to simplify implementation of the present invention
Example is open and helps to understand one or more of the various inventive aspects, above to the exemplary embodiment of the embodiment of the present invention
Description in, each feature of the embodiment of the present invention is grouped together into single embodiment, figure or descriptions thereof sometimes
In.
However, the disclosed method should not be interpreted as reflecting the following intention: i.e. the claimed invention is implemented
Example requires features more more than feature expressly recited in each claim.More precisely, such as claims institute
As reflection, inventive aspect is all features less than single embodiment disclosed above.Therefore, it then follows specific embodiment party
Thus claims of formula are expressly incorporated in the specific embodiment, wherein each claim itself is real as the present invention
Apply the separate embodiments of example.
Finally, it should be noted that above embodiments are only to illustrate the technical solution of the embodiment of the present invention, rather than it is limited
System；Although the embodiment of the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art it is understood that
It is still possible to modify the technical solutions described in the foregoing embodiments, or part of technical characteristic is carried out etc.
With replacement；And these are modified or replaceed, each embodiment skill of the embodiment of the present invention that it does not separate the essence of the corresponding technical solution
The spirit and scope of art scheme.
Claims (10)
1. a kind of extracting method of eye fundus image medium vessels based on multi-feature fusion characterized by comprising
Based on eye fundus image to be processed, using multiple and different classifiers, multiple and different features of optical fundus blood vessel are extracted respectively, and right
The multiple different characteristic carries out Fusion Features；
Based on the comprehensive characteristics after Fusion Features, the denseCRF model completed using training obtains point of the optical fundus blood vessel
Cut image；
Morphological analysis processing is carried out to the segmented image, extracts the bianry image of the optical fundus blood vessel.
2. the method according to claim 1, wherein the multiple and different features for extracting optical fundus blood vessel respectively
Step specifically includes:
From the eye fundus image to be processed, the edge feature, textural characteristics and spatial form of the optical fundus blood vessel are extracted respectively
Feature.
3. according to the method described in claim 2, it is characterized in that, the edge feature for extracting the optical fundus blood vessel respectively,
The step of textural characteristics and spatial form feature, specifically includes:
Using Canny algorithm classification device, the eye fundus image to be processed is analyzed, the edge feature is extracted；Utilize LBP algorithm point
Class device analyzes the eye fundus image to be processed, extracts the textural characteristics；And using FCN algorithm classification device, analysis it is described to
Eye fundus image is handled, the spatial form feature is extracted.
4. according to the method described in claim 2, it is characterized in that, described carry out Fusion Features to the multiple different characteristic
Step specifically includes:
It is based respectively on the edge feature, the textural characteristics and the spatial form feature, extracts feature vector, and will extract
Each feature vector cascaded, constitute multi-feature vector；
Using probabilistic standard algorithm, the multi-feature vector is normalized, and utilizes principal component analysis algorithm,
Multi-feature vector after normalization is handled, realizes the Fusion Features.
5. according to the method described in claim 4, it is characterized in that, described utilize principal component analysis algorithm, after normalization
The step of multi-feature vector is handled, and realizes the Fusion Features specifically includes:
The autocorrelation matrix of multi-feature vector after seeking the normalization, and utilize principal component analysis algorithm to described from phase
It closes matrix to be decomposed, obtains eigenvectors matrix and eigenvalue matrix respectively；
Multi-feature vector after the normalization is subjected to Karhunen-Loeve transformation, obtains transformed eigenvalue matrix, and described in selection
Feature vector corresponding to maximum eigenvalue in transformed eigenvalue matrix, as the comprehensive characteristics.
6. method according to any one of claims 1-5, which is characterized in that described to carry out form to the segmented image
The step of credit analysis is handled, and extracts the bianry image of the optical fundus blood vessel specifically includes:
Gray processing is carried out to the segmented image to negate, and the high cap of form is carried out to the negated image of gray processing and is filtered,
Image after obtaining cap transformation；
Using Threshold Analysis method, binary conversion treatment is carried out to the image after the cap transformation, obtains the two of the optical fundus blood vessel
It is worth image.
7. according to the method described in claim 3, it is characterized in that, utilize Canny algorithm classification device described, analysis it is described to
Before the step of handling eye fundus image, extracting the edge feature, further includes:
A certain number of eye fundus image training samples are obtained, and utilize the eye fundus image training sample, to the basis of foundation
Canny algorithm classification device, basis LBP algorithm classification device, basis FCN algorithm classification device and basis denseCRF model are iterated
Training obtains the Canny algorithm classification device, the LBP algorithm classification device, the FCN algorithm classification device and described
DenseCRF model.
8. a kind of extraction element of eye fundus image medium vessels based on multi-feature fusion characterized by comprising
Comprehensive characteristics extraction module, using multiple and different classifiers, extracts eyeground blood for being based on eye fundus image to be processed respectively
Multiple and different features of pipe, and Fusion Features are carried out to the multiple different characteristic；
Optical fundus blood vessel image segmentation module, for utilizing the denseCRF of training completion based on the comprehensive characteristics after Fusion Features
Model obtains the segmented image of the optical fundus blood vessel；
Output module is extracted, for carrying out morphological analysis processing to the segmented image, extracts the two-value of the optical fundus blood vessel
Image.
9. a kind of electronic equipment characterized by comprising at least one processor, at least one processor, communication interface and total
Line；
The memory, the processor and the communication interface complete mutual communication, the communication by the bus
Interface is also used to the transmission of the information between the electronic equipment and eye fundus image equipment；
The computer program that can be run on the processor is stored in the memory, the processor executes the calculating
When machine program, the method as described in any in claim 1 to 7 is realized.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer instruction is stored up, the computer instruction makes the computer execute the method as described in any in claim 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811418985.4A CN109615614B (en) | 2018-11-26 | 2018-11-26 | Method for extracting blood vessels in fundus image based on multi-feature fusion and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811418985.4A CN109615614B (en) | 2018-11-26 | 2018-11-26 | Method for extracting blood vessels in fundus image based on multi-feature fusion and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109615614A true CN109615614A (en) | 2019-04-12 |
| CN109615614B CN109615614B (en) | 2020-08-18 |
Family
ID=66003908
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811418985.4A Active CN109615614B (en) | 2018-11-26 | 2018-11-26 | Method for extracting blood vessels in fundus image based on multi-feature fusion and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109615614B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009631A (en) * | 2019-04-15 | 2019-07-12 | 唐晓颖 | Vascular quality appraisal procedure, device, equipment and the medium of eye fundus image |
| CN110349170A (en) * | 2019-07-13 | 2019-10-18 | 长春工业大学 | A kind of full connection CRF cascade FCN and K mean value brain tumor partitioning algorithm |
| CN112257791A (en) * | 2020-10-26 | 2021-01-22 | 重庆邮电大学 | Classification method of multi-attribute classification tasks based on CNN and PCA |
| CN112734723A (en) * | 2021-01-08 | 2021-04-30 | 温州医科大学 | Multi-source data-oriented breast tumor image classification prediction method and device |
| CN114220054A (en) * | 2021-12-15 | 2022-03-22 | 北京中科智易科技有限公司 | Method for analyzing tactical action of equipment and synchronously displaying equipment based on equipment bus data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106530283A (en) * | 2016-10-20 | 2017-03-22 | 北京工业大学 | SVM (support vector machine)-based medical image blood vessel recognition method |
| CN107248161A (en) * | 2017-05-11 | 2017-10-13 | 江西理工大学 | Retinal vessel extracting method is supervised in a kind of having for multiple features fusion |
| CN107704886A (en) * | 2017-10-20 | 2018-02-16 | 北京工业大学 | A kind of medical image hierarchy system and method based on depth convolutional neural networks |
| US20180204094A1 (en) * | 2015-11-26 | 2018-07-19 | Tencent Technology (Shenzhen) Company Limited | Image recognition method and apparatus |
-
2018
- 2018-11-26 CN CN201811418985.4A patent/CN109615614B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180204094A1 (en) * | 2015-11-26 | 2018-07-19 | Tencent Technology (Shenzhen) Company Limited | Image recognition method and apparatus |
| CN106530283A (en) * | 2016-10-20 | 2017-03-22 | 北京工业大学 | SVM (support vector machine)-based medical image blood vessel recognition method |
| CN107248161A (en) * | 2017-05-11 | 2017-10-13 | 江西理工大学 | Retinal vessel extracting method is supervised in a kind of having for multiple features fusion |
| CN107704886A (en) * | 2017-10-20 | 2018-02-16 | 北京工业大学 | A kind of medical image hierarchy system and method based on depth convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| JI-JIANG YANG等: ""Exploiting ensemble learning for automatic cataract detection and grading"", 《COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009631A (en) * | 2019-04-15 | 2019-07-12 | 唐晓颖 | Vascular quality appraisal procedure, device, equipment and the medium of eye fundus image |
| CN110349170A (en) * | 2019-07-13 | 2019-10-18 | 长春工业大学 | A kind of full connection CRF cascade FCN and K mean value brain tumor partitioning algorithm |
| CN110349170B (en) * | 2019-07-13 | 2022-07-08 | 长春工业大学 | Full-connection CRF cascade FCN and K mean brain tumor segmentation algorithm |
| CN112257791A (en) * | 2020-10-26 | 2021-01-22 | 重庆邮电大学 | Classification method of multi-attribute classification tasks based on CNN and PCA |
| CN112734723A (en) * | 2021-01-08 | 2021-04-30 | 温州医科大学 | Multi-source data-oriented breast tumor image classification prediction method and device |
| CN112734723B (en) * | 2021-01-08 | 2023-06-30 | 温州医科大学 | Multi-source data-oriented breast tumor image classification prediction method and device |
| CN114220054A (en) * | 2021-12-15 | 2022-03-22 | 北京中科智易科技有限公司 | Method for analyzing tactical action of equipment and synchronously displaying equipment based on equipment bus data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109615614B (en) | 2020-08-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109615614A (en) | The extracting method and electronic equipment of eye fundus image medium vessels based on multi-feature fusion | |
| CN109154978B (en) | System and method for detecting plant diseases | |
| CN109102037A (en) | Chinese model training, Chinese image-recognizing method, device, equipment and medium | |
| CN111125416A (en) | Image retrieval method based on multi-feature fusion | |
| CN109784366A (en) | The fine grit classification method, apparatus and electronic equipment of target object | |
| CN108062543A (en) | A kind of face recognition method and device | |
| CN108764358A (en) | A kind of Terahertz image-recognizing method, device, equipment and readable storage medium storing program for executing | |
| Villaruz | Deep convolutional neural network feature extraction for berry trees classification | |
| CN108230330B (en) | Method for quickly segmenting highway pavement and positioning camera | |
| Castellano et al. | Deep convolutional embedding for digitized painting clustering | |
| Zhang | Half century for image segmentation | |
| CN116612307A (en) | Solanaceae disease grade identification method based on transfer learning | |
| Pushpa et al. | Deep learning model for plant species classification using leaf vein features | |
| Kwenda et al. | Machine learning methods for forest image analysis and classification: A survey of the state of the art | |
| Kar | Mastering Computer Vision with TensorFlow 2. x: Build advanced computer vision applications using machine learning and deep learning techniques | |
| Zabihi et al. | Vessel extraction of conjunctival images using LBPs and ANFIS | |
| CN112801238B (en) | Image classification method and device, electronic equipment and storage medium | |
| RU2768544C1 (en) | Method for recognition of text in images of documents | |
| CN115641575A (en) | Leafhopper agricultural pest detection method based on sparse candidate frame | |
| CN113837062A (en) | Classification method and device, storage medium and electronic equipment | |
| Putzu | Computer aided diagnosis algorithms for digital microscopy | |
| Vununu et al. | A Deep Feature Learning Scheme for Counting the Cells in Microscopy Data | |
| Vojjila | Image Segmentation and Classification of Marine Organisms | |
| SELAMAWIT | GABOR FEATURE AND LOCAL FEATURE ASSISTED CNN FOR ETHIOPIAN ORTHODOX TEWAHEDO CHURCH ICON DETECTION | |
| Huttinga | Insights into deep learning methods with application to cancer imaging |
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 |