CN107506770A - Diabetic retinopathy eye-ground photography standard picture generation method - Google Patents
Diabetic retinopathy eye-ground photography standard picture generation method Download PDFInfo
- Publication number
- CN107506770A CN107506770A CN201710706334.4A CN201710706334A CN107506770A CN 107506770 A CN107506770 A CN 107506770A CN 201710706334 A CN201710706334 A CN 201710706334A CN 107506770 A CN107506770 A CN 107506770A
- Authority
- CN
- China
- Prior art keywords
- image
- mrow
- feature
- standard picture
- local
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
A kind of diabetic retinopathy eye-ground photography standard picture generation method, comprises the following steps：Step (1), by the nonstandard eye fundus image of collection pass through generation model generate new samples image；Step (2), the extraction of new samples image local feature；Step (3), nonstandard image local feature and standard picture local feature contrasted in discrimination model, unanimously then export new samples image, that is, the standard picture generated is inconsistent then to adjust new samples image.Method proposed by the present invention is easy and effective, and the standard picture definition of generation reaches intelligent auxiliary diagnosis system requirements, improves rate of correct diagnosis.
Description
Technical field
The present invention relates to a kind of field of medical image processing, more particularly, to a kind of diabetic retinopathy eye-ground photography
Standard picture generation method, for artificial intelligence medical diagnosis.
Background technology
Diabetic retinopathy (Diabetic Retinopathy, DR) is common blinding illness in eye.China is the whole world 2
The most country of diabetes mellitus type, DR illness rate, blind rate also raise year by year, are that current work age groups are primary
Blinding disease.Illness rates of China DR in diabetes suffer from crowd is 24.7%~37.5% at present.According to international glycosuria
Sick alliance's statistical result showed, by 2015, the people of China diabetic about 1.1 hundred million, China DR patient is calculated about by this
27000000 people, with significantly increasing for diabetic, following DR patient populations can be doubled and redoubled, and DR preventing and treating turns into more next
More important social concern.And according to planning commission's statistics is defended, there are 3.2 ten thousand oculists in China at present, wherein, it is engaged in eyeground medical treatment clothes
Business and the doctor about 800-1000 people of research, for the diabetic more than 100,000,000, oculist's wretched insufficiency, lead
Cause China's sugar to net sick examination rate less than 10%, to change this present situation, can be by DR eyeground that artificial intelligence diagnosis clearly mark
Photo solves.
At present, 87% diabetic goes to a doctor at county level and following medical institutions.The DR intelligent auxiliary diagnosis researched and developed
DR eye-ground photographies image used in model training is both from good eye-ground photography equipment and the horizontal doctor of good DR collections
The large-scale Grade III Class A hospital of teacher, image quality are higher.But undertake basic unit's ophthalmology of a large amount of DR examinations tasks, professional equipment and
The scarcity of professional eye doctor, DR eye-ground photographies image definition and camera angle of acquisition etc. is set all to be extremely difficult to DR intelligently auxiliary
Help the perfect condition of diagnosis input requirements.The DR eye fundus images of standard how are generated by non-type DR eye fundus images, are desirable
The problem solved.
Chinese patent (application number：CN201410078378) a kind of eye fundus image feature of diabetic retinopathy is disclosed
Extracting method, this method comprise the following steps：(1) eye fundus image RGB channel selects；(2) eye fundus image optic disk positions；(3) it is right
Eye fundus image carries out hard exudate feature and cotton-wool patches feature extraction, if finding at least one feature, i.e., the eye after generation extraction
Base map picture, if not finding feature, carry out aneurysms feature and inter-retinal hemorrhage feature extraction, the eye after regeneration extraction
Base map picture.
Chinese papers：Vessel extraction method (the author of diabetic retinopathy image：Dan Lingyu, in May, 2015), should
Method includes：(1) colored retinal images are pre-processed, obtains blood vessel and the higher green passage figure of background parts contrast
Picture；(2) lesion detection processing and image enhaucament are carried out respectively to green channel image；(3) lesion detection knot is obtained based on step (2)
Fruit, then lesion region is split；(4) the enhancing image obtained based on step (2), by existing retinal vessel point
Segmentation method is contrasted, then in conjunction with the architectural characteristic of blood vessel --- and it is thinner more toward tip, using only a kind of threshold segmentation method
Minute blood vessel can not be effectively partitioned into, it is proposed that a kind of Segmentation Method of Retinal Blood Vessels based on global threshold and local threshold
It is handled, obtains preliminary vessel segmentation；(5) the hard exudate segmentation result obtained based on step (3), by it
Removed from the vessel segmentation in step (4), then by the blood vessel to being broken in lesion region according to gradient intensity and
The similar principle in direction is attached, and obtains accurate vessel segmentation.
However, it is exactly not no image (the namely nonstandard eye to gathering for the first time that above-mentioned prior art, which has common defects,
Base map picture) secondary operation processing is carried out, intelligent auxiliary diagnosis system requirements is reached with the standard picture definition of generation, raising is examined
The purpose of disconnected accuracy.By this technology, the disturbing factor in nonstandard image can be removed, retains diabetic retinopathy
The characteristics of lesion of eye-ground photography image, for artificial intelligence medical diagnosis.
The content of the invention
For basic hospital professional equipment and the scarcity of professional eye doctor, make the DR eye-ground photography image definitions of acquisition
The problem of being all extremely difficult to the perfect condition of DR intelligent auxiliary diagnosis input requirements with camera angle etc., the present invention will utilize depth
GAN technologies in study, the DR eye-ground photographies image optimization lifting that basic unit's difference Ophthalmologic apparatus is obtained can not remove standard picture
Disturbing factor in standard picture, retain the characteristics of lesion of diabetic retinopathy eye fundus image, reaching DR, intelligently auxiliary is examined
The input picture standard requirement of disconnected system, so as to obtain preferable diagnostic result.Therefore the invention discloses a kind of diabetes to regard
Retinopathy eye-ground photography standard picture generation method, to solve the deficiencies in the prior art.Its technical scheme is as follows：
A kind of diabetic retinopathy eye-ground photography standard picture generation method, comprises the following steps：
Step (1), by the nonstandard eye fundus image of collection pass through generation model generate new samples image；
Step (2), the extraction of new samples image local feature；
Step (3), nonstandard image local feature and standard picture local feature contrasted in discrimination model, unanimously then
New samples image is exported, that is, the standard picture generated is inconsistent then to adjust new samples image.
Beneficial effect：Method proposed by the present invention is easy and effective, and the standard picture definition of generation reaches intelligence auxiliary and examined
Disconnected system requirements, improves rate of correct diagnosis.
Brief description of the drawings
Fig. 1 is without the image handled by the present invention.
Fig. 2 is the standard picture after the inventive method is handled.
Fig. 3 is diabetic retinopathy eye-ground photography standard picture generation method FB(flow block) of the present invention.
Embodiment
A kind of diabetic retinopathy eye-ground photography standard picture generation method, comprises the following steps：
Step (1), by the nonstandard eye fundus image of collection pass through generation model generate new samples image；
Step (2), the extraction of new samples image local feature；The local feature includes local direction histogram of gradients HOG
Feature, Scale invariant features transform SIFT feature, local color features；
Step (3), nonstandard image local feature and standard picture local feature contrasted in discrimination model, unanimously then
New samples image is exported, that is, the standard picture generated is inconsistent then to adjust new samples image.
Embodiment 1
Step (1) further comprises：The generation model used for deep learning in generation confrontation network G AN models, GAN by
Ian Goodfellow proposed that its main thought was in 2014, trained a maker (Generator, abbreviation G), from
Sample true to nature is generated in machine noise or latent variable, while trains an arbiter (Discriminator, abbreviation D)
Differentiate True Data and generation data, both train simultaneously, until reaching a Nash Equilibrium --- the data of maker generation
With authentic specimen indifference, arbiter also can not correctly distinguish generation data and True Data.The model can be by non-type sugar
Urinate disease PVR image generation standard picture；For GAN models, its optimization problem is a minimum-maximization problem, its
Shown in object function such as formula (1)；Wherein x is sampled in True Data distribution pdata(x), z is sampled in prior distribution pz(z) (such as
Gaussian noise distribution), E () represents to calculate desired value.When fixed generation network G, the optimization for differentiating network D,
It is understood that：Input comes from True Data, and D optimizations network structure makes oneself to export 1, and input comes from generation data, D
Optimization network structure makes oneself to export 0；When fixed differentiation network D, G, which optimizes the network of oneself, makes oneself output as far as possible
With the sample as True Data, and cause generation sample by D differentiation after, D output high probability.
Embodiment 2
The extracting method of the local direction histogram of gradients HOG features is as follows：Small connected region is divided the image into first
Domain, i.e.,：Cell factory, gradient or edge the direction histogram of each pixel in cell factory is then gathered, finally this
A little set of histograms constitutive characteristic describer altogether.
The extracting method of the local direction histogram of gradients HOG features is further described below：By a regional area
Image is carried out：
1) gray processing (regarding image as an x, y, z (gray scale) 3-D view)；
2) standardization of color space is carried out to input picture using standardization Gamma spaces and color space correction method
(normalization)；The purpose is to adjust the contrast of image, reduce image local shade and illumination variation caused by influence, together
When can suppress the interference of noise.
In order to reduce the influence of illumination factor, it is necessary first to which whole image is standardized (normalization).In image
In texture strength, the proportion of local top layer exposure contribution is larger, so, this compression processing can be effectively reduced image office
The shade and illumination variation in portion.Because colouring information effect is little, gray-scale map is generally first converted into；Gamma compresses formula：
I (x, y)=I (x, y)gamma
Such as it can be taken as：Gamma=1/2
3) gradient (including size and Orientation) of each pixel of image is calculated；Primarily to capture profile information, simultaneously
The interference that further weakened light shines.The gradient in image abscissa and ordinate direction is calculated, and calculates each location of pixels accordingly
Gradient direction value；Derivation operations can not only capture profile, the shadow and some texture informations, moreover it is possible to what further weakened light shone
Influence.The gradient of pixel (x, y) is in image：
Gx(x, y)=H (x+1, y)-H (x-1, y)
Gy(x, y)=H (x, y+1)-H (x, y-1)
G in formulax(x, y), Gy(x, y), H (x, y) represent the horizontal direction ladder at pixel (x, y) place in input picture respectively
Degree, vertical gradient and pixel value.The gradient magnitude and gradient direction at pixel (x, y) place be respectively：
Its method is as follows：Convolution algorithm is done to original image with [1,0,1] gradient operator first, obtains x directions (level side
To with to the right for positive direction) gradient component gradscalx, convolution fortune is then done to original image with [1,0, -1] T gradient operators
Calculate, obtain the gradient component gradscaly of y directions (vertical direction, with upwards for positive direction).Then calculated again with above formula
The gradient magnitude of the pixel and direction.
4) junior unit (such as 6*6 pixels/unit) is divided an image into；Local area image is divided into several " units
Lattice cell ", the gradient information of this 6*6 pixel is counted using 9 bin histogram.Namely by cell gradient direction
360 degree are divided into 9 direction blocks, and gradient magnitude is exactly the weights as projection.
5) histogram of gradients (numbers of different gradients) of each unit is counted, you can form the descriptor of each unit；
Cell factory is combined into big block (block), normalized gradient histogram in block：
Unit is combined into big, space coconnected piece (blocks).So, in a block all units spy
Sign vector, which is together in series, just obtains the HOG features of the block.These sections are that mutual is overlapping, and this means that：Each unit
The feature of lattice can be appeared in repeatedly in last characteristic vector with different results.We are by the block descriptor after normalization
(vector) just referred to as HOG descriptors.
6) block (such as 3*3 unit/block) will be formed per several units, the feature of all units describes in a block
Symbol, which is together in series, just obtains the HOG feature descriptors of the block.
7) all pieces of HOG feature descriptors in local area image are together in series and can be obtained by the regional area
The HOG feature descriptors of image.This is exactly the final characteristic vector for being available for classification to use.
Embodiment 3
The extracting method of SIFT feature in step 2
SIFT full name is Scale Invariant Feature Transform, Scale invariant features transform, including 4
Individual key step：
1) extremum extracting of metric space：The image searched on all metric spaces, identified by gaussian derivative function
Potentially
To yardstick and select constant point of interest.
Generally usable DoG (difference Gauss, DifferenceoF Gaussina) carry out approximate calculation Laplacian.
If k is the definition of the scale factor, then DoG of two neighboring Gaussian scale-space：
D (x, y, σ)=[G (x, y, k σ)-G (x, y, σ)] * I (x, y)
=L (x, y, k σ)-L (x, y, σ)
Wherein, G (x, y, σ) is gaussian kernel function.σ is referred to as the metric space factor, and it is the standard deviation of Gauss normal distribution,
The degree that image is blurred is reflected, its bigger image of value is fuzzyyer, and corresponding yardstick is also bigger.L (x, y, σ) represents figure
The Gaussian scale-space of picture.
2) bad characteristic point is removed：Extreme point truly is found in discrete space, it is necessary to try that bar will be unsatisfactory for
The point of part weeds out, and is carried out curve fitting by metric space DoG functions and finds extreme point, the essence of this step is to remove DoG
The very asymmetric point of local curvature.
The undesirable point to be weeded out mainly has two kinds：
A. the characteristic point of low contrast
B. unstable skirt response point
Respective handling technology is as follows：
1. reject the characteristic point of low contrast
Candidate feature point x, its offset are defined as Δ x, and its contrast is D (x) absolute value | D (x) |, D (x) is applied
Taylor expansion
Because x is D (x) extreme point, thus to above formula derivation and make its be 0, obtain
Then the Δ x tried to achieve is updated in D (x) Taylor expansion again
If the threshold value of contrast is T, if | D (x^) | >=T, this feature point retain, otherwise weeded out.
2., reject unstable skirt response point
Principal curve value is bigger on the direction of edge gradient, and then principal curve value is smaller along edge direction.Candidate is special
The principal curvatures for levying the DoG function D (x) of point is directly proportional to the characteristic value of 2 × 2Hessian matrix Hs.
Wherein, Dxx, Dxy, DyyIt is that the difference of candidate point neighbor assignment position is tried to achieve.
Only need to detect
Wherein, Tr (H) is the mark of matrix H, and Det (H) is the determinant of matrix H.T γ are threshold value.If above formula is set up,
This feature point is rejected, is otherwise retained.
3) characteristic direction assignment：Gradient direction based on image local, it is one or more to distribute to each key point position
Direction, follow-up all operations are all that line translation is entered in the direction for key point, yardstick and position, so as to provide these features
Consistency.Argument and width centered on characteristic point, using 3 × 1.5 σ as the gradient that each pixel is calculated in the field of radius
Value, is then counted using histogram to the argument of gradient.The transverse axis of histogram is the direction of gradient, and the longitudinal axis is gradient direction
The accumulated value of corresponding gradient magnitude, the direction in histogram corresponding to top are the direction of characteristic point.
The argument and amplitude centered on characteristic point, using 3 × 1.5 σ as the area image of radius are calculated, each point L (x, y)
The mould m (x, y) and direction θ (x, y) of gradient can be tried to achieve by formula below：
4) feature point description measures the local ladder of image in the neighborhood around each characteristic point on selected yardstick
Degree, these gradients are transformed into a kind of expression, this deformation and the light change for representing to allow bigger local shape.
Reference axis is rotated to be to the direction of characteristic point first, the ladder of the pixel of 16 × 16 window centered on characteristic point
Spend amplitude and direction, the pixel in window be divided into 16 pieces, every piece be 8 directions in its pixel statistics with histogram, altogether can shape
Into the characteristic vector of 128 dimensions.
Embodiment 4
The extracting method of local color features in step 2
1) new samples image is traveled through with different size of window, forms various sizes of topography；
2) it is local using color image segmentation method, extraction according to standard picture focus color characteristic in different windows
Image focus color；
3) topography's ratio value shared by calculating foci colored pixels point, it is topography's color characteristic.
Embodiment 5
Discrimination model in step (3) is differentiated using SVM (SVMs)
By the local feature of the new samples image zooming-out of generation compared with standard picture local feature, if its feature gap
Within the specific limits, then it is assumed that be real, otherwise it is assumed that being false.If vacation, then the new samples figure for needing modification to generate
Picture, its feature is extracted again and is differentiated, untill differentiating truly.
Many details are elaborated in the above description in order to fully understand the present invention.But above description is only
Presently preferred embodiments of the present invention, the invention can be embodied in many other ways as described herein, therefore this
Invention is not limited by specific implementation disclosed above.Any those skilled in the art are not departing from the technology of the present invention simultaneously
In the case of aspects, all technical solution of the present invention is made using the methods and technical content of the disclosure above many possible
Changes and modifications, or it is revised as the equivalent embodiment of equivalent variations.Every content without departing from technical solution of the present invention, according to this
The technical spirit of invention still falls within skill of the present invention to any simple modifications, equivalents, and modifications made for any of the above embodiments
In the range of the protection of art scheme.
Claims (8)
1. a kind of diabetic retinopathy eye-ground photography standard picture generation method, it is characterized in that：Comprise the following steps：
Step (1), by the nonstandard eye fundus image of collection pass through generation model generate new samples image；
Step (2), the extraction of new samples image local feature；
Step (3), nonstandard image local feature and standard picture local feature contrasted in discrimination model, unanimously then exported
New samples image, that is, the standard picture generated are inconsistent then to adjust new samples image.
2. a kind of diabetic retinopathy eye-ground photography standard picture generation method according to claim 1, its feature
For：The step (1) further comprises：The generation model used for deep learning in generation confrontation network G AN models, the confrontation
Network G AN models generate standard picture by non-type diabetic retinopathy image；For GAN models, its object function
As shown in formula (1)：
<mrow>
<munder>
<mi>min</mi>
<mi>G</mi>
</munder>
<munder>
<mi>max</mi>
<mi>D</mi>
</munder>
<mi>V</mi>
<mrow>
<mo>(</mo>
<mi>D</mi>
<mo>,</mo>
<mi>G</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msub>
<mi>E</mi>
<mrow>
<mi>x</mi>
<mo>~</mo>
<msub>
<mi>p</mi>
<mrow>
<mi>d</mi>
<mi>a</mi>
<mi>t</mi>
<mi>a</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
</mrow>
</msub>
<mo>&lsqb;</mo>
<mi>log</mi>
<mi> </mi>
<mi>D</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
<mo>+</mo>
<msub>
<mi>E</mi>
<mrow>
<mi>z</mi>
<mo>~</mo>
<msub>
<mi>p</mi>
<mi>z</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
</mrow>
</msub>
<mo>&lsqb;</mo>
<mi>log</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>-</mo>
<mi>D</mi>
<mo>(</mo>
<mrow>
<mi>G</mi>
<mrow>
<mo>(</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein x is sampled in True Data distribution p data (x), and z is sampled in prior distribution pz (z) (such as Gaussian noise distribution), E
() represents to calculate desired value；When fixed generation network G, the optimization for differentiating network D, it will be understood that：It is defeated
Enter and come from True Data, D optimizations network structure makes oneself to export " 1 ", and input comes from generation data, D optimization network structures
Make oneself to export " 0 "；When it is fixed differentiate network D when, G, which optimizes the network of oneself, makes oneself output as far as possible and True Data
The same sample, and cause the sample of generation after D differentiation, D output high probabilities.
3. a kind of diabetic retinopathy eye-ground photography standard picture generation method according to claim 1, its feature
For：The local feature includes local direction histogram of gradients HOG features, Scale invariant features transform SIFT feature, local face
Color characteristic.
4. a kind of diabetic retinopathy eye-ground photography standard picture generation method according to claim 3, its feature
For：The extracting method of the local direction histogram of gradients HOG features is as follows：Small connected region is divided the image into first, i.e.,：
Cell factory, gradient or edge the direction histogram of each pixel in cell factory is then gathered, finally these Nogatas
Figure combines constitutive characteristic describer.
5. a kind of diabetic retinopathy eye-ground photography standard picture generation method according to claim 4, its feature
For：The extracting method of the local direction histogram of gradients HOG features is further described below：One local area image is entered
OK：1) gray processing；2) standard of color space is carried out to input picture using standardization Gamma spaces and color space correction method
Change；3) gradient of each pixel of image is calculated；4) junior unit is divided an image into；It is " single that local area image is divided into several
First lattice cell ", using histogram come the gradient information of statistical pixel；5) histogram of gradients of each unit is counted, you can formed
The descriptor of each unit；6) block will be formed per several units, the feature descriptor of all units is connected in a block
Just to obtain the HOG feature descriptors of the block；7) all pieces of HOG feature descriptors in local area image are together in series
It can be obtained by the HOG feature descriptors of the local area image.
6. a kind of diabetic retinopathy eye-ground photography standard picture generation method according to claim 3, its feature
For：The Scale invariant features transform SIFT feature extracting method comprises the following steps：1) extremum extracting of metric space：Search
Image on all metric spaces, point of interest potentially constant to yardstick and selection is identified by gaussian derivative function；2)
Remove bad characteristic point：Extreme point truly is found in discrete space, it is necessary to try the point rejecting for the condition that is unsatisfactory for
Fall, carried out curve fitting by metric space DoG functions and find extreme point, the essence of this step is that to remove DoG local curvatures non-
Normal asymmetric point；3) characteristic direction assignment：Gradient direction based on image local, distribute to each key point position one or
Multiple directions, follow-up all operations are all that line translation is entered in the direction for key point, yardstick and position, special so as to provide these
The consistency of sign；4) feature point description：In the neighborhood around each characteristic point, the part of image is measured on selected yardstick
Gradient, these gradients are transformed into a kind of expression, this deformation and the light change for representing to allow bigger local shape.
7. a kind of diabetic retinopathy eye-ground photography standard picture generation method according to claim 3, its feature
For：The extracting method of the local color features further comprises following steps：1) new samples are traveled through with different size of window
Image, form various sizes of topography；2) according to standard picture focus color characteristic in different windows, using cromogram
As dividing method, extraction topography focus color；3) topography's ratio value shared by calculating foci colored pixels point, as should
Topography's color characteristic.
8. a kind of diabetic retinopathy eye-ground photography standard picture generation method according to claim 1, its feature
For：Discrimination model in the step (3) differentiated using SVM, by the local feature of the new samples image zooming-out of generation with
Standard picture local feature compares, if its feature gap is within the specific limits, then it is assumed that is real, otherwise it is assumed that being false
's；If vacation, then the new samples image for needing modification to generate, its feature is extracted again and is differentiated, be truly until differentiating
Only.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710706334.4A CN107506770A (en) | 2017-08-17 | 2017-08-17 | Diabetic retinopathy eye-ground photography standard picture generation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710706334.4A CN107506770A (en) | 2017-08-17 | 2017-08-17 | Diabetic retinopathy eye-ground photography standard picture generation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107506770A true CN107506770A (en) | 2017-12-22 |
Family
ID=60692159
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710706334.4A Pending CN107506770A (en) | 2017-08-17 | 2017-08-17 | Diabetic retinopathy eye-ground photography standard picture generation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107506770A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108510503A (en) * | 2018-03-19 | 2018-09-07 | 哈尔滨工程大学 | A kind of multi-texturing characteristic image dividing method based on Nash Equilibrium Theory |
| CN108537801A (en) * | 2018-03-29 | 2018-09-14 | 山东大学 | Based on the retinal angiomatous image partition method for generating confrontation network |
| CN108765379A (en) * | 2018-05-14 | 2018-11-06 | 深圳明眸科技有限公司 | Computational methods, device, Medical Devices and the storage medium of eyeground pathological changes region area |
| CN108765408A (en) * | 2018-05-31 | 2018-11-06 | 杭州同绘科技有限公司 | Build the method in cancer pathology image virtual case library and the multiple dimensioned cancer detection system based on convolutional neural networks |
| CN109166126A (en) * | 2018-08-13 | 2019-01-08 | 苏州比格威医疗科技有限公司 | A method of paint crackle is divided on ICGA image based on condition production confrontation network |
| CN109300138A (en) * | 2018-08-02 | 2019-02-01 | 哈尔滨工程大学 | A kind of medical image cutting method improving C-V model based on two step Nash Equilibriums |
| CN109409508A (en) * | 2018-11-06 | 2019-03-01 | 成都信息工程大学 | A method of model avalanche is solved based on confrontation Web vector graphic perception loss is generated |
| CN109508647A (en) * | 2018-10-22 | 2019-03-22 | 北京理工大学 | A kind of spectra database extended method based on generation confrontation network |
| CN109934116A (en) * | 2019-02-19 | 2019-06-25 | 华南理工大学 | A kind of standard faces generation method based on generation confrontation mechanism and attention mechanism |
| CN109978796A (en) * | 2019-04-04 | 2019-07-05 | 北京百度网讯科技有限公司 | Optical fundus blood vessel Picture Generation Method, device and storage medium |
| CN110211087A (en) * | 2019-01-28 | 2019-09-06 | 南通大学 | The semi-automatic diabetic eyeground pathological changes mask method that can share |
| CN110222728A (en) * | 2019-05-15 | 2019-09-10 | 图灵深视(南京)科技有限公司 | The training method of article discrimination model, system and article discrimination method, equipment |
| CN110517248A (en) * | 2019-08-27 | 2019-11-29 | 北京百度网讯科技有限公司 | Processing, training method, device and its equipment of eye fundus image |
| CN110674828A (en) * | 2018-07-03 | 2020-01-10 | 江威 | Method and device for normalizing fundus images |
| CN110751187A (en) * | 2019-09-26 | 2020-02-04 | 上海联影智能医疗科技有限公司 | Training method of abnormal area image generation network and related product |
| CN111080524A (en) * | 2019-12-19 | 2020-04-28 | 吉林农业大学 | Plant disease and insect pest identification method based on deep learning |
| CN111325809A (en) * | 2020-02-07 | 2020-06-23 | 广东工业大学 | Appearance image generation method based on double-impedance network |
| CN111971689A (en) * | 2018-04-13 | 2020-11-20 | 医科达有限公司 | Image synthesis using countermeasure networks |
| JP2021079042A (en) * | 2019-11-22 | 2021-05-27 | キヤノン株式会社 | Image processing device, image processing method, and program |
| CN115985472A (en) * | 2022-12-01 | 2023-04-18 | 珠海全一科技有限公司 | Fundus image labeling method and system based on neural network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106709958A (en) * | 2016-12-03 | 2017-05-24 | 浙江大学 | Gray scale gradient and color histogram-based image quality evaluation method |
| CN106951867A (en) * | 2017-03-22 | 2017-07-14 | 成都擎天树科技有限公司 | Face identification method, device, system and equipment based on convolutional neural networks |
-
2017
- 2017-08-17 CN CN201710706334.4A patent/CN107506770A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106709958A (en) * | 2016-12-03 | 2017-05-24 | 浙江大学 | Gray scale gradient and color histogram-based image quality evaluation method |
| CN106951867A (en) * | 2017-03-22 | 2017-07-14 | 成都擎天树科技有限公司 | Face identification method, device, system and equipment based on convolutional neural networks |
Non-Patent Citations (5)
| Title |
|---|
| BROOK_ICV: "《SIFT概述》", 《HTTPS://WWW.CNBLOGS.COM/WANGGUCHANGQING/P/4853263.HTML》 * |
| IAN J. GOODFELLOW等: "《Generative Adversarial Nets》", 《ARXIV:1406.2661 [CS, STAT]》 * |
| RANJIEWEN: "《HOG：用于人体检测的梯度方向直方图 Histograms of Oriented Gradients for Human Detection》", 《HTTPS://WWW.CNBLOGS.COM/RANJIEWEN/ARTICLES/6820541.HTML》 * |
| 王坤峰等: "《生成式对抗网络GAN的研究进展与展望》", 《自动化学报》 * |
| 郭晓新等: "《视网膜血流造影图像序列的超分辨率重建》", 《吉林大学学报（理学版）》 * |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108510503B (en) * | 2018-03-19 | 2021-05-14 | 哈尔滨工程大学 | Multi-texture feature image segmentation method based on Nash equilibrium theory |
| CN108510503A (en) * | 2018-03-19 | 2018-09-07 | 哈尔滨工程大学 | A kind of multi-texturing characteristic image dividing method based on Nash Equilibrium Theory |
| CN108537801A (en) * | 2018-03-29 | 2018-09-14 | 山东大学 | Based on the retinal angiomatous image partition method for generating confrontation network |
| CN111971689A (en) * | 2018-04-13 | 2020-11-20 | 医科达有限公司 | Image synthesis using countermeasure networks |
| CN108765379B (en) * | 2018-05-14 | 2019-11-19 | 深圳明眸科技有限公司 | Calculation method, device, Medical Devices and the storage medium of eyeground pathological changes region area |
| CN108765379A (en) * | 2018-05-14 | 2018-11-06 | 深圳明眸科技有限公司 | Computational methods, device, Medical Devices and the storage medium of eyeground pathological changes region area |
| CN108765408A (en) * | 2018-05-31 | 2018-11-06 | 杭州同绘科技有限公司 | Build the method in cancer pathology image virtual case library and the multiple dimensioned cancer detection system based on convolutional neural networks |
| CN110674828A (en) * | 2018-07-03 | 2020-01-10 | 江威 | Method and device for normalizing fundus images |
| CN109300138A (en) * | 2018-08-02 | 2019-02-01 | 哈尔滨工程大学 | A kind of medical image cutting method improving C-V model based on two step Nash Equilibriums |
| CN109166126B (en) * | 2018-08-13 | 2022-02-18 | 苏州比格威医疗科技有限公司 | Method for segmenting paint cracks on ICGA image based on condition generation type countermeasure network |
| CN109166126A (en) * | 2018-08-13 | 2019-01-08 | 苏州比格威医疗科技有限公司 | A method of paint crackle is divided on ICGA image based on condition production confrontation network |
| CN109508647A (en) * | 2018-10-22 | 2019-03-22 | 北京理工大学 | A kind of spectra database extended method based on generation confrontation network |
| CN109409508B (en) * | 2018-11-06 | 2022-03-15 | 成都信息工程大学 | Method for solving model collapse based on generation of confrontation network use perception loss |
| CN109409508A (en) * | 2018-11-06 | 2019-03-01 | 成都信息工程大学 | A method of model avalanche is solved based on confrontation Web vector graphic perception loss is generated |
| CN110211087B (en) * | 2019-01-28 | 2023-06-30 | 南通大学 | Sharable semiautomatic marking method for diabetic fundus lesions |
| CN110211087A (en) * | 2019-01-28 | 2019-09-06 | 南通大学 | The semi-automatic diabetic eyeground pathological changes mask method that can share |
| CN109934116A (en) * | 2019-02-19 | 2019-06-25 | 华南理工大学 | A kind of standard faces generation method based on generation confrontation mechanism and attention mechanism |
| CN109978796B (en) * | 2019-04-04 | 2021-06-01 | 北京百度网讯科技有限公司 | Fundus blood vessel picture generation method and device and storage medium |
| CN109978796A (en) * | 2019-04-04 | 2019-07-05 | 北京百度网讯科技有限公司 | Optical fundus blood vessel Picture Generation Method, device and storage medium |
| CN110222728B (en) * | 2019-05-15 | 2021-03-12 | 图灵深视(南京)科技有限公司 | Training method and system of article identification model and article identification method and equipment |
| CN110222728A (en) * | 2019-05-15 | 2019-09-10 | 图灵深视(南京)科技有限公司 | The training method of article discrimination model, system and article discrimination method, equipment |
| CN110517248A (en) * | 2019-08-27 | 2019-11-29 | 北京百度网讯科技有限公司 | Processing, training method, device and its equipment of eye fundus image |
| CN110751187B (en) * | 2019-09-26 | 2022-08-19 | 武汉中科医疗科技工业技术研究院有限公司 | Training method of abnormal area image generation network and related product |
| CN110751187A (en) * | 2019-09-26 | 2020-02-04 | 上海联影智能医疗科技有限公司 | Training method of abnormal area image generation network and related product |
| JP7254682B2 (en) | 2019-11-22 | 2023-04-10 | キヤノン株式会社 | Image processing device, image processing method, and program |
| JP2021079042A (en) * | 2019-11-22 | 2021-05-27 | キヤノン株式会社 | Image processing device, image processing method, and program |
| CN111080524A (en) * | 2019-12-19 | 2020-04-28 | 吉林农业大学 | Plant disease and insect pest identification method based on deep learning |
| CN111325809A (en) * | 2020-02-07 | 2020-06-23 | 广东工业大学 | Appearance image generation method based on double-impedance network |
| CN111325809B (en) * | 2020-02-07 | 2021-03-12 | 广东工业大学 | Appearance image generation method based on double-impedance network |
| CN115985472A (en) * | 2022-12-01 | 2023-04-18 | 珠海全一科技有限公司 | Fundus image labeling method and system based on neural network |
| CN115985472B (en) * | 2022-12-01 | 2023-09-22 | 珠海全一科技有限公司 | Fundus image labeling method and fundus image labeling system based on neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107506770A (en) | Diabetic retinopathy eye-ground photography standard picture generation method | |
| CN106408564B (en) | A kind of method for processing fundus images based on deep learning, apparatus and system | |
| CN105513077B (en) | A kind of system for diabetic retinopathy screening | |
| CN107369160A (en) | A kind of OCT image median nexus film new vessels partitioning algorithm | |
| CN109635846A (en) | A kind of multiclass medical image judgment method and system | |
| US9684959B2 (en) | Methods and systems for automatic location of optic structures in an image of an eye, and for automatic retina cup-to-disc ratio computation | |
| CN109858540B (en) | Medical image recognition system and method based on multi-mode fusion | |
| CN106530283A (en) | SVM (support vector machine)-based medical image blood vessel recognition method | |
| CN107644418B (en) | Optic disk detection method and system based on convolutional neural networks | |
| CN100373397C (en) | Pre-processing method for iris image | |
| CN107330449A (en) | A kind of BDR sign detection method and device | |
| CN108765392B (en) | Digestive tract endoscope lesion detection and identification method based on sliding window | |
| Harangi et al. | Automatic detection of the optic disc using majority voting in a collection of optic disc detectors | |
| US20150125052A1 (en) | Drusen lesion image detection system | |
| CN112967285B (en) | Chloasma image recognition method, system and device based on deep learning | |
| CN107977682A (en) | Lymph class cell sorting method and its device based on the enhancing of polar coordinate transform data | |
| CN108615239A (en) | Tongue image dividing method based on threshold technology and Gray Projection | |
| CN109035227A (en) | The system that lung tumors detection and diagnosis is carried out to CT image | |
| CN107563996A (en) | A kind of new discus nervi optici dividing method and system | |
| CN102567734A (en) | Specific value based retina thin blood vessel segmentation method | |
| CN110705468A (en) | Eye movement range identification method and system based on image analysis | |
| Hao et al. | Anterior chamber angles classification in anterior segment OCT images via multi-scale regions convolutional neural networks | |
| CN103279960B (en) | A kind of image partition method of human body cache based on X-ray backscatter images | |
| CN106960199A (en) | A kind of RGB eye is as the complete extraction method in figure white of the eye region | |
| Choudhary et al. | ANN glaucoma detection using cup-to-disk ratio and neuroretinal rim |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171222 |