CN109993756A - A kind of general medical image cutting method based on graph model Yu continuous successive optimization - Google Patents
A kind of general medical image cutting method based on graph model Yu continuous successive optimization Download PDFInfo
- Publication number
- CN109993756A CN109993756A CN201910281453.9A CN201910281453A CN109993756A CN 109993756 A CN109993756 A CN 109993756A CN 201910281453 A CN201910281453 A CN 201910281453A CN 109993756 A CN109993756 A CN 109993756A
- Authority
- CN
- China
- Prior art keywords
- medical image
- image
- function
- relaxation
- continuous
- 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
- 238000005457 optimization Methods 0.000 title claims abstract description 44
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000005520 cutting process Methods 0.000 title claims abstract description 16
- 230000011218 segmentation Effects 0.000 claims abstract description 23
- 238000003709 image segmentation Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000002599 functional magnetic resonance imaging Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 6
- 239000003814 drug Substances 0.000 claims description 5
- 238000009826 distribution Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000003384 imaging method Methods 0.000 claims description 3
- 238000003860 storage Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000012512 characterization method Methods 0.000 claims 2
- 230000000694 effects Effects 0.000 abstract description 7
- 230000008859 change Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 210000004072 lung Anatomy 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 210000005166 vasculature Anatomy 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
Abstract
The invention discloses a kind of general medical image cutting method based on graph model Yu continuous successive optimization, more particularly to field of medical image processing, six steps are exported including new medical image input, image preprocessing, serialization relaxation, the successive optimization based on convex-concave relaxation and segmentation result based on graph model building Second-Order Discrete objective function, dispersive target function.The present invention by using graph model and continuous successive optimization solution route, medical image segmentation can be automatically performed by not depending on mark sample, towards general medical image segmentation task, it can handle a plurality of types of medical images such as CT, PET, fMRI, using continuous successive optimization method, medical image segmentation efficiency and effect have been taken into account, can greatly have been improved efficiency in medical image segmentation batch processing task and precision.
Description
Technical field
The present invention relates to field of medical image processing, it is more particularly related to a kind of based on graph model and continuous
The general medical image cutting method of successive optimization.
Background technique
In big data era, the scale of medical image (CT image, PET image, fMRI image etc.) also shows magnanimity increasing
Long situation is automatically analyzed for the medical image expansion batch of magnanimity, is the weight of the medical image research under area of computer aided
Want direction.
Medical image segmentation is the important link and basic operation of Medical Image Processing, for accurately describing interested solution
Structure or tissue are cutd open, is handled convenient for the quantization of junior scholar's subsequent statistical with intellectual analysis, from the point of view of algorithm angle, it refers to according to certain
Medical image is split into the process of similar, the mutually independent subregion of multiple internal properties, at this stage medicine figure by optiaml ciriterion
As common method such as Threshold segmentation, region segmentation method, the classifier methods of dividing method, since mixing is blocked in medical image
Target, fuzzy boundary, significant noise, structure or tissue changeability and other unknown causes, point of medical image
Cutting is still a challenging problem.
Threshold method and region method are more suitable for the segmentation with certain uniformity in the above method, and classifier side
Method generally requires the sample largely marked.Different from these methods, the medical image point minimized based on graph model and energy
Segmentation method is introduced into smoothness constraint and obtains excellent effect in a plurality of types of medical image segmentation tasks, while not needing mark sample
This.But, the existing method minimized based on graph model and energy is often limited to specific energy function form (secondary mould letter
Number) or specific graph structure.
Summary of the invention
In order to overcome the drawbacks described above of the prior art, the embodiment of the present invention provide it is a kind of based on graph model and it is continuous gradually
The general medical image cutting method of optimization does not depend on mark by using the solution route of graph model and continuous successive optimization
Sample can be automatically performed medical image segmentation, towards general medical image segmentation task, can handle CT, PET, fMRI etc.
A plurality of types of medical images have taken into account medical image segmentation efficiency and effect using continuous successive optimization method, in medicine figure
As that can greatly be improved efficiency and precision in segmentation batch processing task.
To achieve the above object, the invention provides the following technical scheme: it is a kind of based on graph model and continuous successive optimization
General medical image cutting method, specific step is as follows for dividing method:
S1: new medical image input；
S1.1: inputting medical image in computer port, a new picture is transmitted every time, until all picture transfers
It finishes；
S1.2: medical image type is arranged according to input image information, and determines region number determination, initialization segmentation is appointed
Business；
S1.3: identification medical image storage format obtains single channel or multi-channel medical image by codec format；
S2: image preprocessing；
S2.1: carrying out denoising to the medical image of input, and reducing error in imaging process influences；
S2.2: according to image mission requirements to be processed, image enhancement is carried out to the image after denoising, highlights pass as far as possible
The key area of note；
S3: Second-Order Discrete objective function is constructed based on graph model；
S3.1: using the digraph with Markov property, i.e. figure Markov random field, image is indicated, wherein image
In each pixel by vertex representation each in digraph, and construct neighbor relationships；
S3.2: according to the maximum a posteriori probability model of graph model building image multi-tag segmentation in S3.1；
S3.3: logarithmic transformation is carried out to maximum a posteriori probability model, objective function is changed into Second-Order Discrete majorized function；
S4: the serialization relaxation of dispersive target function；
S4.1: solving Second-Order Discrete optimization problem in S3.3 using the method for continuous successive optimization, and Second-Order Discrete is excellent
Change problem relaxation be a continuous optimization problems, continuous solution is gradually then mapped as discrete solution, by relations of distribution matrix X by from
It dissipates domain Π and is relaxed to continuous domain Ω；
S5: the successive optimization based on convex-concave relaxation；
S5.1: the convex relaxation function F of building objective function F (X)vex(X) with recessed relaxation function Fcave(X), make convex loose letter
Several and recessed relaxation function is identical as former objective function value in discrete domain, i.e.,
F (X)=Fvex(X)=Fcave(X), X ∈ Π
S5.2: convex-concave relaxation objective function F is constructed using convex combinationη(X), specific as follows；
max Fη(X)=η Fvex(X)+(1-η)Fcave(X),
S.t.X ∈ Ω, η ∈ [0,1]
Wherein η is convex combination coefficient；
S5.3: by the way that η is gradually reduced to 0 from 1, the objective function that convex-concave relaxation solves gradually easily is solved from one
Convex optimization problem be changed into the recessed relaxation problem with former optimization problem equivalence, thus gradually automatically derive one preferably from
Dissipate solution X；
S6: segmentation result output；
S6.1: using segmentation tag matrix X is obtained to each vertex, i.e., each pixel in image distributes cut zone
Label, and visualize and obtain image segmentation result；
S6.2: judging whether batch processing terminates, and according to judging result determine task continues and return and again divide or
Person terminates and exits.
In a preferred embodiment, the step 1 traditional Chinese medicine input object image is set as CT image, PET figure
Picture or fMRI image.
In a preferred embodiment, maximum a posteriori probability model is specially in the S3.2
Max P (L | Q)=P (Q | L) P (L)
The wherein degree of agreement of the segmentation tag of likelihood item P (Q | L) characterization pixel and picture appearance information, priori item P (L)
Characterize the smooth relationship between neighbour.
In a preferred embodiment, logarithmic transformation is carried out to maximum a posteriori model in the S3.3, by target letter
Number is changed into Second-Order Discrete majorized function, and optimization problem is changed at this time
Max F (X)=D (X)+S (X),
s.t.X∈Π.
Wherein each vertex in figure, that is, each pixel and segmentation tag in image the relations of distribution by X ∈ 0,
1}N×LIt indicates, wherein N indicates that vertex number, L indicate number of tags, and Π is discrete set, be may be defined as
Data item D (X) is an order polynomial function, and smooth item S (X) is quadratic polynomial function.
In a preferred embodiment, in the S4.1, wherein continuous domain is defined as
In a preferred embodiment, in the S5.1, the value in continuous domain Ω is respectively convex function and concave function,
In addition, optimal solution and former objective function optimal solution in discrete domain Π of the recessed relaxation function in continuous domain Ω is of equal value.
Technical effect and advantage of the invention:
1, by using the solution route of graph model and continuous successive optimization, doctor can be automatically performed by not depending on mark sample
It learns image segmentation and can handle a plurality of types of medicine figures such as CT, PET, fMRI towards general medical image segmentation task
Picture；
2, by using continuous successive optimization method, medical image segmentation efficiency and effect have been taken into account, in medical image point
Cutting can greatly improve efficiency and precision does not have for medical image feature flexible design objective function in batch processing task
The constraint of the functional forms such as submodular function.
Detailed description of the invention
Fig. 1 is technical solution of the present invention overall flow figure.
Fig. 2 is successive optimization flow chart of the invention.
Fig. 3 is the contrast effect figure of original image and segmentation figure under the method for the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
A kind of general medical image cutting method based on graph model Yu continuous successive optimization according to shown in Fig. 1-3, point
Specific step is as follows for segmentation method:
S1: new medical image input, medical image input object are set as CT image, PET image or fMRI image, face
To general medical image segmentation task, it can handle a plurality of types of medical images such as CT, PET, fMRI；
S1.1: inputting medical image in computer port, a new picture is transmitted every time, until all picture transfers
It finishes；
S1.2: medical image type is arranged according to input image information, and determines region number determination, initialization segmentation is appointed
Business；
S1.3: identification medical image storage format obtains single channel or multi-channel medical image by codec format；
S2: image preprocessing；
S2.1: carrying out denoising to the medical image of input, and reducing error in imaging process influences；
S2.2: according to image mission requirements to be processed, image enhancement is carried out to the image after denoising, highlights pass as far as possible
The key area of note, such as the vasculature part of lung CT image is enhanced conducive to morphological operation；
S3: Second-Order Discrete objective function is constructed based on graph model；
S3.1: using the digraph with Markov property, i.e. figure Markov random field, image is indicated, wherein image
In each pixel by vertex representation each in digraph, and construct neighbor relationships；
S3.2: according to the maximum a posteriori probability model of graph model building image multi-tag segmentation in S3.1, specially
Max P (L | Q)=P (Q | L) P (L)
The wherein degree of agreement of the segmentation tag of likelihood item P (Q | L) characterization pixel and picture appearance information, priori item P (L)
Characterize the smooth relationship between neighbour；
S3.3: carrying out logarithmic transformation to maximum a posteriori probability model, objective function be changed into Second-Order Discrete majorized function,
Logarithmic transformation is carried out to maximum a posteriori model, objective function is changed into Second-Order Discrete majorized function, optimization problem changes at this time
For
Max F (X)=D (X)+S (X),
s.t.X∈Π.
Wherein each vertex in figure, that is, each pixel and segmentation tag in image the relations of distribution by X ∈ 0,
1}N×LIt indicates, wherein N indicates that vertex number, L indicate number of tags, and Π is discrete set, be may be defined as
Data item D (X) is an order polynomial function, and smooth item S (X) is quadratic polynomial function, is based on different from other
The medical image cutting method of optimization, such as figure segmentation method (using energy minimization models), this method is for second order term without secondary
The additional constraints such as modular function；
S4: the serialization relaxation of dispersive target function；
S4.1: solving Second-Order Discrete optimization problem in S3.3 using the method for continuous successive optimization, and Second-Order Discrete is excellent
Change problem relaxation be a continuous optimization problems, continuous solution is gradually then mapped as discrete solution, by relations of distribution matrix X by from
It dissipates domain Π and is relaxed to continuous domain Ω, continuous domain is defined as
S5: the successive optimization based on convex-concave relaxation；
S5.1: the convex relaxation function F of building objective function F (X)vex(X) with recessed relaxation function Fcave(X), make convex loose letter
Several and recessed relaxation function is identical as former objective function value in discrete domain, i.e.,
F (X)=Fvex(X)=Fcave(X), X ∈ Π
Value in continuous domain Ω is respectively convex function and concave function, in addition, optimal solution of the recessed relaxation function in continuous domain Ω
Optimal solution with former objective function in discrete domain Π is of equal value, by using continuous successive optimization method, has taken into account medical image point
Efficiency and effect are cut, can greatly be improved efficiency in medical image segmentation batch processing task and precision；
S5.2: convex-concave relaxation objective function F is constructed using convex combinationη(X), specific as follows；
max Fη(X)=η Fvex(X)+(1-η)Fcave(X),
S.t.X ∈ Ω, η ∈ [0,1]
Wherein η is convex combination coefficient；
S5.3: by the way that η is gradually reduced to 0 from 1, the objective function that convex-concave relaxation solves gradually easily is solved from one
Convex optimization problem be changed into the recessed relaxation problem with former optimization problem equivalence, thus gradually automatically derive one preferably from
Dissipate solution X；
S6: segmentation result output；
S6.1: using segmentation tag matrix X is obtained to each vertex, i.e., each pixel in image distributes cut zone
Label, and visualize and obtain image segmentation result；
S6.2: judging whether batch processing terminates, and according to judging result determine task continues and return and again divide or
Person terminates and exits, and specially if batch processing is not finished, reads new medical image, returns to step 1；If batch processing knot
Beam, then task terminates and exits.
The several points that should finally illustrate are: firstly, the present invention discloses in embodiment attached drawing, relating only to and the embodiment of the present disclosure
The structure being related to, other structures, which can refer to, to be commonly designed, under not conflict situations, the same embodiment of the present invention and different implementations
Example can be combined with each other；
Last: the foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, all in the present invention
Spirit and principle within, any modification, equivalent replacement, improvement and so on, should be included in protection scope of the present invention it
It is interior.
Claims (6)
1. a kind of general medical image cutting method based on graph model Yu continuous successive optimization, it is characterised in that: dividing method
Specific step is as follows:
S1: new medical image input；
S1.1: inputting medical image in computer port, a new picture is transmitted every time, until all picture transfers are complete
Finish；
S1.2: medical image type is arranged according to input image information, and determines region number determination, initializes segmentation task；
S1.3: identification medical image storage format obtains single channel or multi-channel medical image by codec format；
S2: image preprocessing；
S2.1: carrying out denoising to the medical image of input, and reducing error in imaging process influences；
S2.2: according to image mission requirements to be processed, image enhancement is carried out to the image after denoising, highlights concern as far as possible
Key area；
S3: Second-Order Discrete objective function is constructed based on graph model；
S3.1: using the digraph with Markov property, i.e. figure Markov random field, image is indicated, wherein in image
Each pixel constructs neighbor relationships by vertex representation each in digraph；
S3.2: according to the maximum a posteriori probability model of graph model building image multi-tag segmentation in S3.1；
S3.3: logarithmic transformation is carried out to maximum a posteriori probability model, objective function is changed into Second-Order Discrete majorized function；
S4: the serialization relaxation of dispersive target function；
S4.1: solving Second-Order Discrete optimization problem in S3.3 using the method for continuous successive optimization, and Second-Order Discrete optimization is asked
Topic relaxation is a continuous optimization problems, continuous solution is gradually then mapped as discrete solution, by relations of distribution matrix X by discrete domain
Π is relaxed to continuous domain Ω；
S5: the successive optimization based on convex-concave relaxation；
S5.1: the convex relaxation function F of building objective function F (X)vex(X) with recessed relaxation function Fcave(X), make convex relaxation function with
Recessed relaxation function is identical as former objective function value in discrete domain, i.e.,
F (X)=Fvex(X)=Fcave(X), X ∈ Π
S5.2: convex-concave relaxation objective function F is constructed using convex combinationη(X), specific as follows；
max Fη(X)=η Fvex(X)+(1-η)Fcave(X),
S.t.X ∈ Ω, η ∈ [0,1]
Wherein η is convex combination coefficient；
S5.3: by the way that η is gradually reduced to 0 from 1, the objective function that convex-concave relaxation solves gradually easily solves from one convex
Optimization problem is changed into the recessed relaxation problem with former optimization problem equivalence, thus gradually automatically derive a preferable discrete solution
X；
S6: segmentation result output；
S6.1: using segmentation tag matrix X is obtained to each vertex, i.e., each pixel in image distributes cut zone label,
And it visualizes and obtains image segmentation result；
S6.2: judging whether batch processing terminates, and is continued according to judging result decision task and return and divide again or tie
Beam simultaneously exits.
2. a kind of general medical image cutting method based on graph model Yu continuous successive optimization according to claim 1,
It is characterized by: the step 1 traditional Chinese medicine input object image is set as CT image, PET image or fMRI image.
3. a kind of general medical image cutting method based on graph model Yu continuous successive optimization according to claim 1,
It is characterized by: maximum a posteriori probability model is specially in the S3.2
Max P (L | Q)=P (Q | L) P (L)
The wherein degree of agreement of the segmentation tag of likelihood item P (Q | L) characterization pixel and picture appearance information, priori item P (L) characterization
Smooth relationship between neighbour.
4. a kind of general medical image cutting method based on graph model Yu continuous successive optimization according to claim 1,
It is characterized by: carrying out logarithmic transformation to maximum a posteriori model in the S3.3, objective function is changed into Second-Order Discrete optimization
Function, optimization problem is changed at this time
Max F (X)=D (X)+S (X),
s.t.X∈Π.
Wherein each vertex in figure, that is, each pixel and segmentation tag in image the relations of distribution by X ∈ { 0,1 }N×L
It indicates, wherein N indicates that vertex number, L indicate number of tags, and Π is discrete set, be may be defined as
Data item D (X) is an order polynomial function, and smooth item S (X) is quadratic polynomial function.
5. a kind of general medical image cutting method based on graph model Yu continuous successive optimization according to claim 1,
It is characterized by: wherein continuous domain is defined as in the S4.1
6. a kind of general medical image cutting method based on graph model Yu continuous successive optimization according to claim 1,
It is characterized by: the value in continuous domain Ω is respectively convex function and concave function in the S5.1, in addition, recessed relaxation function is even
The optimal solution of the optimal solution of continuous domain Ω and former objective function in discrete domain Π is of equal value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910281453.9A CN109993756B (en) | 2019-04-09 | 2019-04-09 | General medical image segmentation method based on graph model and continuous stepwise optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910281453.9A CN109993756B (en) | 2019-04-09 | 2019-04-09 | General medical image segmentation method based on graph model and continuous stepwise optimization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109993756A true CN109993756A (en) | 2019-07-09 |
| CN109993756B CN109993756B (en) | 2022-04-15 |
Family
ID=67131039
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910281453.9A Active CN109993756B (en) | 2019-04-09 | 2019-04-09 | General medical image segmentation method based on graph model and continuous stepwise optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109993756B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115758899A (en) * | 2021-06-28 | 2023-03-07 | 山东华科信息技术有限公司 | Transformer fault diagnosis system based on graph Markov neural network |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040267529A1 (en) * | 2003-06-24 | 2004-12-30 | Aurilab, Llc | N-gram spotting followed by matching continuation tree forward and backward from a spotted n-gram |
| US20110241927A1 (en) * | 2010-04-02 | 2011-10-06 | Porikli Fatih M | Method for Detecting Small Targets in Radar Images Using Needle Based Hypotheses Verification |
| CN102509105A (en) * | 2011-09-30 | 2012-06-20 | 北京航空航天大学 | Hierarchical processing method of image scene based on Bayesian inference |
| CN104778683A (en) * | 2015-01-27 | 2015-07-15 | 杭州电子科技大学 | Multi-modal image segmenting method based on functional mapping |
| CN105335524A (en) * | 2015-11-27 | 2016-02-17 | 中国科学院自动化研究所 | Graph search algorithm applied to large-scale irregular structure data |
| CN106709921A (en) * | 2016-12-29 | 2017-05-24 | 华侨大学 | Color image segmentation method based on space Dirichlet hybrid model |
| CN108629815A (en) * | 2017-03-15 | 2018-10-09 | 北京京东尚科信息技术有限公司 | image generating method and device |
| CN109741378A (en) * | 2018-12-13 | 2019-05-10 | 华南理工大学 | Multimodal medical image registration method, apparatus, platform and medium based on MRF model |
-
2019
- 2019-04-09 CN CN201910281453.9A patent/CN109993756B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040267529A1 (en) * | 2003-06-24 | 2004-12-30 | Aurilab, Llc | N-gram spotting followed by matching continuation tree forward and backward from a spotted n-gram |
| US20110241927A1 (en) * | 2010-04-02 | 2011-10-06 | Porikli Fatih M | Method for Detecting Small Targets in Radar Images Using Needle Based Hypotheses Verification |
| CN102509105A (en) * | 2011-09-30 | 2012-06-20 | 北京航空航天大学 | Hierarchical processing method of image scene based on Bayesian inference |
| CN104778683A (en) * | 2015-01-27 | 2015-07-15 | 杭州电子科技大学 | Multi-modal image segmenting method based on functional mapping |
| CN105335524A (en) * | 2015-11-27 | 2016-02-17 | 中国科学院自动化研究所 | Graph search algorithm applied to large-scale irregular structure data |
| CN106709921A (en) * | 2016-12-29 | 2017-05-24 | 华侨大学 | Color image segmentation method based on space Dirichlet hybrid model |
| CN108629815A (en) * | 2017-03-15 | 2018-10-09 | 北京京东尚科信息技术有限公司 | image generating method and device |
| CN109741378A (en) * | 2018-12-13 | 2019-05-10 | 华南理工大学 | Multimodal medical image registration method, apparatus, platform and medium based on MRF model |
Non-Patent Citations (5)
| Title |
|---|
| HUO, Q等: "BAYESIAN ADAPTIVE LEARNING OF THE PARAMETERS OF HIDDEN MARKOV MODEL FOR SPEECH RECOGNITION", 《IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING》 * |
| 刘丽: "近重复文本图像匹配研究", 《中国博士学位论文全文数据库》 * |
| 李晶等: "基于渐非凸渐凹化过程的子图匹配算法", 《北京航空航天大学学报》 * |
| 臧顺全: "基于图割优化的Markov随机场图像分割方法综述", 《电视技术》 * |
| 辛芳芳等: "利用小波域HMC模型进行遥感图像变化检测", 《西安电子科技大学学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115758899A (en) * | 2021-06-28 | 2023-03-07 | 山东华科信息技术有限公司 | Transformer fault diagnosis system based on graph Markov neural network |
| CN115758899B (en) * | 2021-06-28 | 2023-05-09 | 山东华科信息技术有限公司 | Transformer fault diagnosis system based on graph Markov neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109993756B (en) | 2022-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jiang et al. | A novel negative-transfer-resistant fuzzy clustering model with a shared cross-domain transfer latent space and its application to brain CT image segmentation | |
| Belacel | Multicriteria assignment method PROAFTN: Methodology and medical application | |
| Ajala Funmilola et al. | Fuzzy kc-means clustering algorithm for medical image segmentation | |
| US8224089B2 (en) | Information processing device and information processing method, image processing device and image processing method, and computer program | |
| Fang et al. | Crop leaf disease grade identification based on an improved convolutional neural network | |
| Figueira et al. | Electre Tri-C, a multiple criteria decision aiding sorting model applied to assisted reproduction | |
| Sadafi et al. | Multiclass deep active learning for detecting red blood cell subtypes in brightfield microscopy | |
| Seetharaman et al. | Texture characterization, representation, description, and classification based on full range Gaussian Markov random field model with Bayesian approach | |
| Huang et al. | Deep PET/CT fusion with Dempster-Shafer theory for lymphoma segmentation | |
| US20220245802A1 (en) | Systems and methods for classifying biomedical image data using a graph neural network | |
| Cerrolaza et al. | Fetal skull segmentation in 3D ultrasound via structured geodesic random forest | |
| Popordanoska et al. | On the relationship between calibrated predictors and unbiased volume estimation | |
| CN109993756A (en) | A kind of general medical image cutting method based on graph model Yu continuous successive optimization | |
| Yang et al. | Cranial implant prediction by learning an ensemble of slice-based skull completion networks | |
| Salsabili et al. | Fully automated estimation of the mean linear intercept in histopathology images of mouse lung tissue | |
| Mortaheb et al. | Metal artifact reduction and segmentation of dental computerized tomography images using least square support vector machine and mean shift algorithm | |
| US7996340B2 (en) | Method and system for L1-based robust distribution clustering of multinomial distributions | |
| Hu et al. | Multi-class medical image segmentation using one-vs-rest graph cuts and majority voting | |
| Shen et al. | An ILP model for multi-label MRFs with connectivity constraints | |
| US8725724B2 (en) | Method for efficient association of multiple distributions | |
| Pérez-Carrasco et al. | Segmentation of bone structures in 3D CT images based on continuous max-flow optimization | |
| Hassanien et al. | Enhanced rough sets rule reduction algorithm for classification digital mammography | |
| Vasylieva et al. | Automation Methods for Processing Medical Images Based on the Application of Grids. | |
| Borovec et al. | Region growing using superpixels with learned shape prior | |
| Kromp et al. | Machine learning framework incorporating expert knowledge in tissue image annotation |
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 |