CN103366379B - Level set medical image partition method based on heredity Kernel fuzzy clustering - Google Patents
Level set medical image partition method based on heredity Kernel fuzzy clustering Download PDFInfo
- Publication number
- CN103366379B CN103366379B CN201310321955.2A CN201310321955A CN103366379B CN 103366379 B CN103366379 B CN 103366379B CN 201310321955 A CN201310321955 A CN 201310321955A CN 103366379 B CN103366379 B CN 103366379B
- Authority
- CN
- China
- Prior art keywords
- function
- image
- colony
- medical image
- cluster
- 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.)
- Active
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Level set medical image partition method based on heredity Kernel fuzzy clustering, relates to the application of medical image segmentation.The present invention utilizes heredity Kernel fuzzy clustering algorithm to obtain the optimum cluster result of pending medical image, then cluster result is applied to the initial profile of LBF model and splits image, can accomplish have higher segmentation efficiency and accuracy to blood-vessel image.
Description
Technical field
The invention belongs to image processing field, relate to the application of medical image segmentation.It is specifically related to genetic algorithm (GA),
Fuzzy c-means clustering algorithm (KFCM) and the Level Set Method application in image segmentation field.
Background technology
Medical image segmentation, as an important component part in image segmentation field, is constantly subjected to international academic community
Paying attention to, a large amount of scholars obtain notable achievement in this field.But, for same clinical image, due to different application purposes, sense
The tissue of interest will be different.So, select the most according to demand suitable partitioning algorithm be medical image segmentation field so far
An insoluble difficult problem.Such as, for the MR image of same width brain, can be divided into according to different demands: carrying of cerebral tissue
Take, the classification of cerebral tissue, the extraction etc. of specific part cerebral tissue structure, the dividing method between this three differs greatly.Therefore,
A kind of unified dividing method is not the most also had can different types of medical image effectively to be split.
Recent decades, Level Set Method has obtained considerable house show in image segmentation field, and the method was in quilt in 1988
Osher and Sethian proposes at first.It is bent that closed curve evolution problem is converted into level set function in the high one-dimensional space by the method
Line develop implicit mode solve, it can preferably process image topology change problem so that segmentation result with
The topological structure of initial curve is unrelated, thus is widely used in image segmentation field.
At present, image Segmentation Technology Application comparison model widely based on level set has: peripheral type parted pattern, region
Property parted pattern, shape prior type parted pattern and mixed type parted pattern.Peripheral type parted pattern mainly utilizes the ladder of image
Degree information terminates curve evolvement, i.e. promotes evolution curve to approach objective contour by local edge information.Due to edge pattern
Type is not the most applied merely with the marginal information of image, the area information for image, so for there is no obvious gradient variable
This model segmentation effect of weak-edge image changed is the best, and in order to solve this problem, region based segmentation model is suggested, this model
The area information utilizing image promotes evolution curve to approach to objective contour, and most typical domain type model is that han and Vese exists
The C-V Level Set Models proposed on the basis of Level Set Method, this model utilizes evolution inside or outside of curve portion gray average to promote
Curve evolvement.Shape prior pattern type utilizes level set function to express the shape of sample, directly develops level set function
The main component of curve is analyzed, then by the training of target shape is obtained target prior shape, then living
Adding shape constraining item in dynamic skeleton pattern, the difficult point of this model is to express accurately objective trait.Mixed type parted pattern
Mainly by former three be combined with each other, use local and the global information of image.
Although achieving immense success based on level set image segmentation area research personnel, but yet suffer from many asking
Topic.Although C-V model applies the area information of image, but it cannot obtain satisfaction to the image that intensity profile is uneven
Segmentation result.To this end, Li and his team propose local binary matching (LBF) model, this model comes by introducing kernel function
Definition local binary matching amount.LBF model can effectively split the image that intensity profile is uneven, and has higher precision,
But, owing to this model have employed the gradient information of image border, thus its that initial profile is chosen comparison is sensitive, and along with
The choosing difference and may cause the far from each other of segmentation result of initial profile.Its antimierophonic performance is the highest, for structure simultaneously
Complicated medical image segmentation DeGrain.
Summary of the invention
Present invention aims to the weak point of above-mentioned LBF model, propose a kind of based on heredity Kernel fuzzy clustering
Medical image cutting method, to solve the problem that LBF model is sensitive for initial profile and noise robustness is low, thus improve
The segmentation quality of medical image.
The technical scheme is that: first the gray value obtaining the dendrogram picture of medical image is clustered, then by each
Cluster centreBeing encoded into character string with 8 bits, composition clusters colony, calculates the object function of cluster respectivelyWith
Ideal adaptation degree, the colony when former generation is used condition, condition is satisfied then terminates algorithm, obtains optimum
Cluster centre, and calculate optimum membership function by below equation,
Utilize maximum membership grade principleDetermine the class belonging to medical image gradation data collection
Not, optimum cluster image is obtained;
Pass through againDetermine the initial profile of Level Set Models;
Finally, it is applied to LBF model medical image is split.
The present invention utilizes heredity Kernel fuzzy clustering algorithm to obtain the optimum cluster result of pending medical image, then poly-
Class result is applied to the initial profile of LBF model and splits image, can accomplish to have blood-vessel image higher segmentation
Efficiency and accuracy.
Concrete grammar comprises the following steps:
1) the dendrogram picture of medical image is obtained:
If former space sampleBy core nonlinear mapping to a feature space, available:
The dot product of the most former sample space at feature space can be by Mercer kernel representation:
Kernel function takes gaussian kernel function;
1-1) set initialization cluster centre number, and according to cluster numbers, the gray value of medical image is clustered, will be each
Individual cluster centreBeing encoded into character string with 8 bits, composition clusters colony;
1-2) set the maximum evolutionary generation of genetic algorithm, group size, crossover probability, mutation probability, core
Parameter, weighted index, stochastic generationIndividual initial population, and make;
1-3) willThe string of binary characters decoding of individual colony, represents with real number form, utilizes below equation to calculate respectively
The object function of clusterWith ideal adaptation degree, whereinIt isThe cluster centre of class,For this center accordingly
The picture of nuclear space:With
And have
1-4) colony when former generation is used condition, terminate algorithm when meeting condition, and skip next
Step, directly carries out step 1-6), and obtain Optimal cluster centers, calculate optimum membership function by below equation;If no
MeetThen carry out next step 1-5):
1-5) when the most maximum evolutionary generationTime, order, utilize roulette dish method, single-point interior extrapolation method, variation to calculate
Current group is selected, intersects by son, mutation operation, produces colony of a new generation, retain when former generation fitness is big simultaneously
Individuality arriveIn, return to step 1-3) again cluster;
Utilize genetic operator that colony is selected, intersect and mutation operation, produce colony of a new generation,
Wherein selection opertor design is as follows:
A) operation is selected: be genetic to the individuality of a new generation according to the cumulative selection of selected probability;Individual selected
Probability is:
WhereinRepresent that in colony, certain is individual,For the size of colony,For ideal adaptation degree;
B) intersection operation: in the character string of coding, one cross point is set by crossover probability, then exchanges two below
Position to character string;
C) mutation operation: the probability arranging variation is 5%, randomly chooses a son touring, and inserts it into one at random
Position;
Utilize maximum membership grade principleDetermine the class belonging to medical image gradation data collection
Not, optimum cluster image is obtained;
2) pass throughDetermine the initial profile of Level Set Models;
3) initial profile obtained according to step 2, is applied to LBF model and splits medical image;
If given image I:Being given artwork image field, d is image dimension;Using curve C as zero
Level set function, regionWithCorrespondence respectivelyWith,
LBF model is represented by:
WhereinIt is weight coefficient,It is respectively regionWithFigure
As the approximation of gray scale, and evolution rule is:
It is gaussian kernel function, Heaviside function, i.e.
It addition, add length regularization termEnsure the slickness of contour curve,
Simultaneously for odd function not loss of gloss slip horizontal after ensureing a period of time, increase energy penalty term:
Gradient descent flow and the calculus of variations is used to obtain following curve evolvement equation:
WhereinIt is Dirac function,WithIt is defined as:
To EVOLUTION EQUATIONDiscretization, obtains following expression:
New level set function is tried to achieve according to above formula, byPositive and negative obtain new cut zone, and judge
Whether level set function restrains, without then forwarding the new level set function of step 2 and cut zone continuation iteration to, no
Then stop iteration, obtain final segmentation result.
Accompanying drawing explanation
Fig. 1 is blood vessel original image.
Fig. 2 is initial profile 1 figure of LBF model.
Fig. 3 is the LBF model segmentation result figure for initial profile 1.
Fig. 4 is initial profile 2 figure of LBF model.
Fig. 5 is the LBF model segmentation result figure for initial profile 2.
Fig. 6 is the segmentation result figure of the present invention.
Fig. 7 is brain MR original image.
Fig. 8 is the brain MR image adding noise.
Fig. 9 is the segmentation result figure of LBF model.
Figure 10 is the segmentation result figure of the present invention.
Detailed description of the invention
The effect of the present invention is further illustrated by following emulation experiment:
Experimental situation is Matlab7.1, CORE i3 CPU, internal memory 4GB.Experimental data is angiographic image and MR figure
Picture.
Implement process as follows:
Step 1: obtain the dendrogram picture of medical image.
Assume former space sampleBy core nonlinear mapping to a feature space, available:
The dot product of the most former sample space at feature space can be by Mercer kernel representation:
Kernel function takes gaussian kernel function herein。
1-1 sets and initializes cluster centre number, and according to cluster numbers, the gray value of medical image is clustered,
By each cluster centreBeing encoded into character string with 8 bits, composition clusters colony.
1-2 sets the maximum evolutionary generation of genetic algorithm, group size, crossover probability, mutation probability, core
Parameter, weighted index, stochastic generationIndividual initial population, and make。
1-3 willThe string of binary characters decoding of individual colony, represents with real number form, utilizes
With
Calculate the object function of cluster respectivelyWith ideal adaptation degree, whereinIt isThe cluster centre of class,
For this center at the picture of corresponding nuclear space, and have
1-4 uses condition to the colony when former generation, condition is satisfied then terminates algorithm, obtains optimum poly-
Class center, and pass through
Calculate optimum membership function, and go to (6), otherwise go to (5).
1-5 is when the most maximum evolutionary generationTime, order, utilize roulette dish method, single-point interior extrapolation method, variation to calculate
Current group is selected, intersects by son, mutation operation, produces colony of a new generation, retain when former generation fitness is big simultaneously
Individuality arriveIn, go to (3).
Utilize genetic operator that colony is selected, intersect and mutation operation, produce colony of a new generation,
Genetic operator is and how to carry out selecting, intersect and making a variation.Wherein selection opertor design is as follows:
(1) operation is selected
Use roulette wheel selection method to select the higher individuality of fitness and be genetic to the next generation, i.e. according to selected probability
Cumulative selection be genetic to a new generation individuality.Individual selected probability is:
WhereinRepresent that in colony, certain is individual,For the size of colony,For ideal adaptation degree.
(2) intersection operation
The operation that intersects uses single-point interior extrapolation method to operate two pairing individualities, i.e. by intersecting in the character string of coding
Probability arranges a cross point, then exchanges the position of two pairs of character strings below.
(3) mutation operation
The probability arranging variation is 5%, randomly chooses a son touring, and inserts it into a random position.
1-6 utilizes maximum membership grade principleDetermine belonging to medical image gradation data collection
Classification, obtains optimum cluster image。
Step 2: pass throughDetermine the initial profile of Level Set Models.
Step 3: the initial profile obtained according to step 2, is applied to LBF model and splits medical image.
Assume given image I:Being given artwork image field, d is image dimension.Using curve C as zero
Level set function, regionWithCorrespondence respectivelyWith, LBF mould
Type is represented by:
Wherein、It is weight coefficient,、It is respectively regionWithFigure
As the approximation of gray scale, and evolution rule is:
It is gaussian kernel function, Heaviside function, i.e.
It addition, add length regularization termEnsure the slickness of contour curve, be simultaneously
Horizontal odd function not loss of gloss slip after guarantee a period of time, increases energy penalty term:
Gradient descent flow and the calculus of variations is used to obtain following curve evolvement equation:
WhereinIt is Dirac function,WithIt is defined as:
。
To EVOLUTION EQUATIONDiscretization, obtains following expression:
New level set function is tried to achieve according to above formula, byPositive and negative obtain new cut zone, and judge water
Whether flat set function restrains, without then forwarding the new level set function of step 2 and cut zone continuation iteration to, otherwise
Stop iteration, obtain final segmentation result.
Experiment 1. is for blood-vessel image model and the comparison of LBF model herein:
The LBF model carrying out initial profile 1 from blood vessel original image Fig. 1 intercepts as shown in Figure 2.LBF model as seen from Figure 3
For the segmentation result of initial profile 1, Fig. 5 is the segmentation result of the initial profile 2 of the LBF model to Fig. 4.
The present invention uses above experimental data to be a widthBlood vessel gray level image, can obtain according to step one
The cluster result of heredity Kernel fuzzy clustering algorithm.In step one, cluster centre is, greatest iteration number, window is big
Little N is, randomly select initial population number, crossover probability, mutation probability, nuclear parameter,
Weighted index.Calculate the object function of cluster respectively according to step 1-3 after Initial parameter setsAnd ideal adaptation
Degree.Existing colony is judged according to step 1-4, wherein, when meetingTime forward to step
Rapid 1-6, finally gives optimum cluster, goes to step 1-5, and the most maximum evolutionary generation when the conditions set forth above are not met
Time, order, utilizing roulette dish method, single-point interior extrapolation method, current group is selected, intersects by mutation operator, make a variation behaviour
Make, produce colony of a new generation, retain the individuality when former generation fitness is big simultaneously and arriveIn, go to (3).More than according to
Step may finally obtain the optimum cluster result of blood-vessel image.Then pass throughDetermine water
The initial profile of flat collection model, hereinIt is that Dick draws function.Final step is the initial profile utilizing heredity kernel clustering to obtain
Substitute into level set function, carry out EVOLUTIONARY COMPUTATION according to step 3, time step in this step, Heaviside function
's, finally give segmentation result Fig. 6 herein.
By experiment, and comparison diagram 6 and Fig. 5 is it can be seen that the segmentation knot of the different LBF models chosen along with initial profile
The most far from each other, segmentation error rate is higher, chooses comparison for initial profile sensitive.And model is not at the beginning of owing to relying on herein
Beginning profile, does not the most exist initial profile sensitive issue, and can be seen that model can be accomplished herein from experimental result
Blood-vessel image is had higher segmentation efficiency and accuracy.
Experiment 2. is for MR image model and the comparison of LBF model herein.
Experimental data is to add Gaussian noiseMR image.Fig. 8 is 0 for adding average in the figure 7, side
Difference is the MR image of 0.02.In step one, cluster centre is, greatest iteration number, window size N is,
Randomly select initial population number, crossover probability, mutation probability, nuclear parameter, power
Weight index.Calculate the object function of cluster respectively according to step 1-3 after Initial parameter setsWith ideal adaptation degree.Existing colony is judged according to step 1-4, wherein, when meetingTime forward step to
1-6, finally gives optimum cluster, goes to step 1-5, and the most maximum evolutionary generation when the conditions set forth above are not met
Time, order, utilizing roulette dish method, single-point interior extrapolation method, current group is selected, intersects by mutation operator, make a variation behaviour
Make, produce colony of a new generation, retain the individuality when former generation fitness is big simultaneously and arriveIn, go to (3).More than according to
Step may finally obtain the optimum cluster result of blood-vessel image.Then pass throughDetermine
The initial profile of Level Set Models, hereinIt is that Dick draws function.Final step is the initial wheel utilizing heredity kernel clustering to obtain
Wide substitution level set function, carries out EVOLUTIONARY COMPUTATION according to step 3, time step in this step, Heaviside letter
Number, finally give segmentation result Figure 10 herein.
And the segmentation result figure that Fig. 9 is LBF model.
By experiment, comparison diagram 9,10 it can be seen that LBF model is more sensitive for noise ratio, and invention algorithm phase herein
Relatively LBF model has more preferable segmentation precision and efficiency, has higher robustness.Wherein LBF model average calculation times
For 2.25s, and invention required time is 0.58s herein, greatly enhances computational efficiency.
Claims (1)
1. level set medical image partition method based on heredity Kernel fuzzy clustering, first by obtain the dendrogram of medical image as
Gray value clusters, then by each cluster centreBeing encoded into character string with 8 bits, composition clusters colony, respectively
Calculate the object function of clusterWith ideal adaptation degree, the colony when former generation is used condition
, condition is satisfied then terminates algorithm, obtains Optimal cluster centers, and calculates optimum person in servitude by below equation
Genus degree function,
Utilize maximum membership grade principleDetermine the classification belonging to medical image gradation data collection,
To optimum cluster image;
Pass through againDetermine the initial profile of Level Set Models;
Finally, it is applied to LBF model medical image is split;
It is characterized in that comprising the following steps:
1) the dendrogram picture of medical image is obtained:
If former space sampleBy core nonlinear mapping to a feature space, available:
The dot product of the most former sample space at feature space can be by Mercer kernel representation:
Kernel function takes gaussian kernel function;
1-1) set initialization cluster centre number, and according to cluster numbers, the gray value of medical image is clustered, each is gathered
Class centerBeing encoded into character string with 8 bits, composition clusters colony;
1-2) set the maximum evolutionary generation of genetic algorithm, group size, crossover probability, mutation probability, nuclear parameter
, weighted index, stochastic generationIndividual initial population, and make;
1-3) willThe string of binary characters decoding of individual colony, represents with real number form, utilizes below equation to calculate cluster respectively
Object functionWith ideal adaptation degree, whereinIt isThe cluster centre of class,For this center at corresponding nuclear space
Picture:With
And have
1-4) colony when former generation is used condition, terminate algorithm when meeting condition, and skip next step,
Directly carry out step 1-6), and obtain Optimal cluster centers, calculate optimum membership function by below equation;If being unsatisfactory forThen carry out next step 1-5);
1-5) when the most maximum evolutionary generationTime, order, utilize roulette dish method, single-point interior extrapolation method, mutation operator pair
Current group carries out selecting, intersects, mutation operation, produces colony of a new generation, simultaneously retain when former generation fitness big
Body arrivesIn, return to step 1-3) again cluster;
Utilize genetic operator that colony is selected, intersect and mutation operation, produce colony of a new generation,
Wherein selection opertor design is as follows:
A) operation is selected: be genetic to the individuality of a new generation according to the cumulative selection of selected probability;Individual selected probability
For:
WhereinRepresent that in colony, certain is individual,For the size of colony,For ideal adaptation degree;
B) intersection operation: in the character string of coding, one cross point is set by crossover probability, then exchanges two pairs of words below
The position of symbol string;
C) mutation operation: the probability arranging variation is 5%, randomly chooses a son touring, and inserts it into a random position
Put;
1-6) utilize maximum membership grade principleDetermine the class belonging to medical image gradation data collection
Not, optimum cluster image is obtained;
2) pass throughDetermine the initial profile of Level Set Models;
3) according to step 2) initial profile that obtains, it is applied to LBF model and medical image is split;
If given image I:,Being given artwork image field, d is image dimension;Using curve C as zero level
Set function, regionWithCorrespondence respectivelyWith,
LBF model is represented by:
Wherein、It is weight coefficient,WithIt is respectively regionWith
The approximation of gradation of image, and evolution rule is:
It is gaussian kernel function, Heaviside function, i.e.
It addition, add length regularization termEnsure the slickness of contour curve, simultaneously
For odd function not loss of gloss slip horizontal after ensureing a period of time, increase energy penalty term:
Gradient descent flow and the calculus of variations is used to obtain following curve evolvement equation:
WhereinIt is Dirac function,WithIt is defined as:
To EVOLUTION EQUATIONDiscretization, obtains following expression:
New level set function is tried to achieve according to above formula, byPositive and negative obtain new cut zone, and determined level collection
Whether function restrains, and without then forwarding the new level set function of step 2 and cut zone continuation iteration to, otherwise stops
Iteration, obtains final segmentation result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310321955.2A CN103366379B (en) | 2013-07-29 | 2013-07-29 | Level set medical image partition method based on heredity Kernel fuzzy clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310321955.2A CN103366379B (en) | 2013-07-29 | 2013-07-29 | Level set medical image partition method based on heredity Kernel fuzzy clustering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103366379A CN103366379A (en) | 2013-10-23 |
CN103366379B true CN103366379B (en) | 2016-08-10 |
Family
ID=49367647
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310321955.2A Active CN103366379B (en) | 2013-07-29 | 2013-07-29 | Level set medical image partition method based on heredity Kernel fuzzy clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103366379B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103942799B (en) * | 2014-04-25 | 2017-02-01 | 哈尔滨医科大学 | Breast ultrasounography image segmentation method and system |
CN104616308A (en) * | 2015-02-12 | 2015-05-13 | 大连民族学院 | Multiscale level set image segmenting method based on kernel fuzzy clustering |
CN107567637B (en) * | 2015-04-30 | 2022-01-25 | 皇家飞利浦有限公司 | Brain tissue classification |
CN106127734B (en) * | 2016-06-13 | 2019-01-11 | 西安电子科技大学 | MRI image dividing method based on obscure idea and level set frame |
CN106504239B (en) * | 2016-10-25 | 2019-06-21 | 南通大学 | A kind of method of liver area in extraction ultrasound image |
CN108182684B (en) * | 2017-12-22 | 2021-06-25 | 河南师范大学 | Image segmentation method and device based on weighted kernel function fuzzy clustering |
CN110648340B (en) * | 2019-09-29 | 2023-03-17 | 惠州学院 | Method and device for processing image based on binary system and level set |
CN111027546B (en) * | 2019-12-05 | 2024-03-26 | 嘉楠明芯(北京)科技有限公司 | Character segmentation method, device and computer readable storage medium |
CN112686916B (en) * | 2020-12-28 | 2024-04-05 | 淮阴工学院 | Curved surface reconstruction system based on heterogeneous multi-region CT scanning data processing |
CN112907526B (en) * | 2021-02-07 | 2022-04-19 | 电子科技大学 | LBF-based satellite telescope lens surface defect detection method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101719277A (en) * | 2009-12-31 | 2010-06-02 | 华中科技大学 | Method for partitioning genetic fuzzy clustering image |
CN101980298A (en) * | 2010-11-04 | 2011-02-23 | 西安电子科技大学 | Multi-agent genetic clustering algorithm-based image segmentation method |
-
2013
- 2013-07-29 CN CN201310321955.2A patent/CN103366379B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101719277A (en) * | 2009-12-31 | 2010-06-02 | 华中科技大学 | Method for partitioning genetic fuzzy clustering image |
CN101980298A (en) * | 2010-11-04 | 2011-02-23 | 西安电子科技大学 | Multi-agent genetic clustering algorithm-based image segmentation method |
Also Published As
Publication number | Publication date |
---|---|
CN103366379A (en) | 2013-10-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103366379B (en) | Level set medical image partition method based on heredity Kernel fuzzy clustering | |
CN105512289B (en) | Image search method based on deep learning and Hash | |
CN110969626B (en) | Method for extracting hippocampus of human brain nuclear magnetic resonance image based on 3D neural network | |
WO2020172838A1 (en) | Image classification method for improvement of auxiliary classifier gan | |
CN106156793A (en) | Extract in conjunction with further feature and the classification method of medical image of shallow-layer feature extraction | |
CN108304826A (en) | Facial expression recognizing method based on convolutional neural networks | |
CN106529574B (en) | Image classification method based on sparse autocoder and support vector machines | |
CN109903396A (en) | A kind of tooth three-dimensional model automatic division method based on surface parameterization | |
CN107506761A (en) | Brain image dividing method and system based on notable inquiry learning convolutional neural networks | |
CN107203999A (en) | A kind of skin lens image automatic division method based on full convolutional neural networks | |
CN107169504B (en) | A kind of hand-written character recognition method based on extension Non-linear Kernel residual error network | |
CN109740652A (en) | A kind of pathological image classification method and computer equipment | |
CN109685819A (en) | A kind of three-dimensional medical image segmentation method based on feature enhancing | |
CN109063724A (en) | A kind of enhanced production confrontation network and target sample recognition methods | |
CN107229904A (en) | A kind of object detection and recognition method based on deep learning | |
CN108985217A (en) | A kind of traffic sign recognition method and system based on deep space network | |
JP2019185665A (en) | Three-dimensional point cloud label learning device, three-dimensional point cloud label estimation device, three-dimensional point cloud label learning method, three-dimensional point cloud label estimation method, and program | |
CN107392224A (en) | A kind of crop disease recognizer based on triple channel convolutional neural networks | |
CN101625755A (en) | Image division method based on watershed-quantum evolution clustering algorithm | |
CN104103093B (en) | Three-dimensional grid semantic marking method based on deep convolution neural network | |
CN110675411A (en) | Cervical squamous intraepithelial lesion recognition algorithm based on deep learning | |
CN103593674B (en) | A kind of cervical lymph node ultrasonoscopy feature selection method | |
CN107203988B (en) | A kind of method and its application for rebuilding three-dimensional volumetric image by two dimensional x-ray image | |
CN109902564A (en) | A kind of accident detection method based on the sparse autoencoder network of structural similarity | |
CN103020979B (en) | Image segmentation method based on sparse genetic clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |