SlideShare a Scribd company logo
1 of 46
Download to read offline
Grey-Level Co-occurrence
features for salt texture
classification
Igor Orlov
Master’s thesis presentation
University of Oslo, Department of Informatics
Supervisor: Anne H Schistad Solberg
Content
• Introduction
• Problem statement and motivation
• Background
• Theoretical review of methods used in thesis
• Implementation
• Input data set, finding optimal GLCM parameters
• Evaluation
• Finding the best classifier, testing on training and test images
• Summary
Motivation
• Salt structures are closely related to carbohydrates discovery
• Segmentation of salt textures on seismic images is needed to build a
correct velocity model
• Time-consuming process, needs automation
Problem statement
Need to segment salt textures on seismic images using grey-level co-
occurrence matrices (GLCMs)
Two main goals:
• Provide an algorithm and visualization tool for finding the optimal
GLCM parameters, such as offset, direction, window size
• Check whether distance matrix between centers of two classes can be
used as a GLCM feature, if yes – estimate performance
Seismic data acquisition
• Initial velocity model is based on
mapping of textures to known
classes
• Then updated iteratively
• In areas of salt it is hard to build
a velocity model, mapping is
done manually or semi-aided on
each step
Cuts of marine seismic images
Inline slice (vertical cut) Timeslice (horizontal cut)
Data is stored in a 3D cube, can be cut in 2 directions:
Salt texture mapping done manually
Textures in seismic images - inline
• Sub-horizontal reflectors (yellow)
• Up-dipping and down-dipping
reflectors (orange)
• Salt structures (blue)
Textures in seismic images - timeslice
• Sub-horizontal reflectors
(yellow)
• Steep-dipping reflectors (orange)
• Salt structures (blue)
Not considered in this thesis
Background for research methods
• Grey-level co-occurrence matrices – GLCM
• GLCM features
• Distance metrics
• Gaussian classifier
Grey-level co-occurrence matrix
• Texture description method –
intensity change histogram
• Matrix where element (x,y)
represents a probability of changing
a pixel intensity from x to y in a local
window of size W when shifting in
direction θ and offset d
• Second order statistic
• Used 5 directions – 0°,45°,90°,135°
and isotropic
GLCM features
• Contrast – intensity contrast in a local window, 0 for const
• Correlation
• Energy – sum squared of GLCM elements
• Homogeneity – closeness of distribution in diagonal
• Entropy, Inertia, Cluster Shade etc.
Distance metrics
• Used to find a distance between class centers
• Euclidean distance
• Mahalanobis – takes into account in-class distribution (Σ – average
covariance matrix), scale invariant
Multivariate Gaussian classifier
• Based on Bayes’ rule
• Assumes Gaussian distribution
• Very simple, trains fast, good
enough
• The likelihood that vector x
belongs to class wj
Part 2. Implementation and evaluation
• Input dataset from North Sea
• Finding optimal GLCM parameters, algorithm and visualization
• Mahalanobis distance matrix
• Training Gaussian classifier
• Estimating classification results on training and test images
Input image and mask
Original image
Mask. Class 1 – sedimentary rocks (sub-
horizontal), class 2 – salt
Masked and normalized
Image pre-processing
• Input image 1401x1501, [-346;399]
• Cut histogram tails, normalize and
shift to [0;255]
• No histogram equalization
Finding the optimal parameters - algorithm
• Choose direction, offset, window size
• For each pixel of class 1, class 2 calculate the GLCM matrix
• Matrix 16x16 can be presented as a feature vector 1x256
• Calculate the average GLCMs for two classes – class centers
• Build a distance matrix where each element is a pixel-wise Mahalanobis
distance between corresponding elements of class center matrices
• Do the same for other direction, offset and window size, compare norm
Mahalanobis distance calculation
• Optimal parameters is where the norm of Mahalanobis distance
matrix between centers of two classes is the greatest
GLCM parameters
• 16 grey levels, fixed. Thus, all GLCMs had size 16x16
• Window size 31x31, 51x51
• 5 directions
• Which direction is expected to be best?
• Offset 1..15 for 31x31, 1..20 for 51x51
Visualization of distance matrices for 51x51
Horizontal Vertical
• Displaying
Mahalanobis distance
between average
GLCMs for two classes
– salt and
sedimentary rocks
Visualization of distance matrices for 51x51
Diagonal Isotropic
Norm of distance matrix, 31x31
Norm of distance matrix, 51x51
Distance matrix for optimal parameters
• Suppress values around the center, enhance values in the middle
and on 3-pixel distance around the center
Own GLCM feature
• Distance GLCM also shows which GLCM elements are more important
for separation of two classes
• Pixel-wise multiplication of GLCM for pixel with Mahalanobis distance
matrix, then sum elements
• Can also be used as a weighting window, to strengthen the GLCM
elements which are more important for separation
• Gives us a new set of features – weighted contrast (contrast of GLCM
weighted with distMatrix), weighted energy etc.
Feature list
• For each pixel of class 1 and class 2 we have calculated the following
ten features:
• Variance
• GLCM contrast + weighted
• GLCM energy + weighted
• GLCM correlation + weighted
• GLCM homogeneity + weighted
• Own GLCM feature – sum of weighted GLCM elements
Gaussian classification
• Trained classifier on one-, two-, three-feature subsets – taken
manually in a random way
• Some where correlated, overtraining
• One feature is not enough
• Tested on original image (can estimate absolute performance) and 2
test images – only visual estimation
Single feature - Contrast
Single feature – distance GLCM (own)
Single feature – Energy (Weighted)
Classification results for single feature
• For training image, class
mapping was provided
• TN – true negative (correctly
classified rocks), FP – false
positive (rocks classified as salt),
TP – true positive (correctly
classified salt).
Contrast and Correlation
• Good TN and TP rate is not a
guarantee of good results
• TN – 94,17%, TP – 91,8%
Energy and Variance
Energy, Own feature and Variance
Contrast, Energy and Own
Contrast, Own and Weighted Energy
Classification results for multiple features
Test images from same dataset
inlinekorr700 inlinekorr780
Classification on test image inlinekorr700
Classification on test image inlinekorr780
Summary - Achieved results
• Found the optimal GLCM parameters for separation of 2 classes:
• Isotropic, offset 2, window 51x51
• Invented and tested own GLCM feature:
• Mahalanobis distance between average GLCMs for two classes
• Trained and tested a set of Gaussian classifiers to find the best feature
combination for salt texture extraction
• Visually best is Contrast, Own GLCM and Weighted Energy
Summary – Disadvantages and drawbacks
• Only dataset from North Sea was investigated
• Maximum dimension of Gaussian classifiers was 3
• Only variance, standard Matlab GLCM features (4) and own GLCM feature were
tested
• Feature selection done manually
• Only visual testing for test images
Summary - Possible improvements
• 3D GLCM – spatial relationship
• Combination with other texture analysis methods – Fourier transform, greater
order statistics etc.
• Own feature – Mahalanobis distance - is based on pixel class mapping. Needs
generalization (formula)
The end
Thanks for attention, questions…

More Related Content

What's hot

Image processing presentataion
Image processing presentataionImage processing presentataion
Image processing presentataionRafi Ullah
 
Image segmentation
Image segmentationImage segmentation
Image segmentationKuppusamy P
 
Lect 03 - first portion
Lect 03 - first portionLect 03 - first portion
Lect 03 - first portionMoe Moe Myint
 
Support Vector Machines
Support Vector MachinesSupport Vector Machines
Support Vector Machinesnextlib
 
Color image processing Presentation
Color image processing PresentationColor image processing Presentation
Color image processing PresentationRevanth Chimmani
 
DIGITAL IMAGE PROCESSING - LECTURE NOTES
DIGITAL IMAGE PROCESSING - LECTURE NOTESDIGITAL IMAGE PROCESSING - LECTURE NOTES
DIGITAL IMAGE PROCESSING - LECTURE NOTESEzhilya venkat
 
IMAGE SEGMENTATION TECHNIQUES
IMAGE SEGMENTATION TECHNIQUESIMAGE SEGMENTATION TECHNIQUES
IMAGE SEGMENTATION TECHNIQUESVicky Kumar
 
Image Filtering in the Frequency Domain
Image Filtering in the Frequency DomainImage Filtering in the Frequency Domain
Image Filtering in the Frequency DomainAmnaakhaan
 
Chapter 6 color image processing
Chapter 6 color image processingChapter 6 color image processing
Chapter 6 color image processingasodariyabhavesh
 
Lec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image SegmentationLec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image SegmentationUlaş Bağcı
 
Application of edge detection
Application of edge detectionApplication of edge detection
Application of edge detectionNaresh Biloniya
 
Content based image retrieval(cbir)
Content based image retrieval(cbir)Content based image retrieval(cbir)
Content based image retrieval(cbir)paddu123
 
Image Enhancement - Point Processing
Image Enhancement - Point ProcessingImage Enhancement - Point Processing
Image Enhancement - Point ProcessingGayathri31093
 
Image texture analysis techniques survey-1
Image texture analysis techniques  survey-1Image texture analysis techniques  survey-1
Image texture analysis techniques survey-1anitadixitjoshi
 

What's hot (20)

Image processing presentataion
Image processing presentataionImage processing presentataion
Image processing presentataion
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
IMAGE SEGMENTATION.
IMAGE SEGMENTATION.IMAGE SEGMENTATION.
IMAGE SEGMENTATION.
 
Lect 03 - first portion
Lect 03 - first portionLect 03 - first portion
Lect 03 - first portion
 
Support Vector Machines
Support Vector MachinesSupport Vector Machines
Support Vector Machines
 
Canny Edge Detection
Canny Edge DetectionCanny Edge Detection
Canny Edge Detection
 
Spatial domain and filtering
Spatial domain and filteringSpatial domain and filtering
Spatial domain and filtering
 
Color image processing Presentation
Color image processing PresentationColor image processing Presentation
Color image processing Presentation
 
DIGITAL IMAGE PROCESSING - LECTURE NOTES
DIGITAL IMAGE PROCESSING - LECTURE NOTESDIGITAL IMAGE PROCESSING - LECTURE NOTES
DIGITAL IMAGE PROCESSING - LECTURE NOTES
 
IMAGE SEGMENTATION TECHNIQUES
IMAGE SEGMENTATION TECHNIQUESIMAGE SEGMENTATION TECHNIQUES
IMAGE SEGMENTATION TECHNIQUES
 
Image Filtering in the Frequency Domain
Image Filtering in the Frequency DomainImage Filtering in the Frequency Domain
Image Filtering in the Frequency Domain
 
Chapter 6 color image processing
Chapter 6 color image processingChapter 6 color image processing
Chapter 6 color image processing
 
Lec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image SegmentationLec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image Segmentation
 
Application of edge detection
Application of edge detectionApplication of edge detection
Application of edge detection
 
SPATIAL FILTER
SPATIAL FILTERSPATIAL FILTER
SPATIAL FILTER
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
EDGE DETECTION
EDGE DETECTIONEDGE DETECTION
EDGE DETECTION
 
Content based image retrieval(cbir)
Content based image retrieval(cbir)Content based image retrieval(cbir)
Content based image retrieval(cbir)
 
Image Enhancement - Point Processing
Image Enhancement - Point ProcessingImage Enhancement - Point Processing
Image Enhancement - Point Processing
 
Image texture analysis techniques survey-1
Image texture analysis techniques  survey-1Image texture analysis techniques  survey-1
Image texture analysis techniques survey-1
 

Viewers also liked

Texture in image processing
Texture in image processing Texture in image processing
Texture in image processing Anna Aquarian
 
Image Texture Analysis
Image Texture AnalysisImage Texture Analysis
Image Texture Analysislalitxp
 
Features image processing and Extaction
Features image processing and ExtactionFeatures image processing and Extaction
Features image processing and ExtactionAli A Jalil
 
Matlab Feature Extraction Using Segmentation And Edge Detection
Matlab Feature Extraction Using Segmentation And Edge DetectionMatlab Feature Extraction Using Segmentation And Edge Detection
Matlab Feature Extraction Using Segmentation And Edge DetectionDataminingTools Inc
 
Improving Performance of Texture Based Face Recognition Systems by Segmenting...
Improving Performance of Texture Based Face Recognition Systems by Segmenting...Improving Performance of Texture Based Face Recognition Systems by Segmenting...
Improving Performance of Texture Based Face Recognition Systems by Segmenting...IDES Editor
 
Low level feature extraction - chapter 4
Low level feature extraction - chapter 4Low level feature extraction - chapter 4
Low level feature extraction - chapter 4Aalaa Khattab
 
Analisis komponen utama (Principal Component Analysis)
Analisis komponen utama (Principal Component Analysis)Analisis komponen utama (Principal Component Analysis)
Analisis komponen utama (Principal Component Analysis)Indah Fitri Hapsari
 
Lbp based edge-texture features for object recoginition
Lbp based edge-texture features for object recoginitionLbp based edge-texture features for object recoginition
Lbp based edge-texture features for object recoginitionIGEEKS TECHNOLOGIES
 
SVM-based CBIR of breast masses on mammograms
SVM-based CBIR of breast masses on mammogramsSVM-based CBIR of breast masses on mammograms
SVM-based CBIR of breast masses on mammogramsLazaros Tsochatzidis
 
04 image enhancement in spatial domain DIP
04 image enhancement in spatial domain DIP04 image enhancement in spatial domain DIP
04 image enhancement in spatial domain DIPbabak danyal
 
Feature Extraction and Principal Component Analysis
Feature Extraction and Principal Component AnalysisFeature Extraction and Principal Component Analysis
Feature Extraction and Principal Component AnalysisSayed Abulhasan Quadri
 
2.6 support vector machines and associative classifiers revised
2.6 support vector machines and associative classifiers revised2.6 support vector machines and associative classifiers revised
2.6 support vector machines and associative classifiers revisedKrish_ver2
 
PPT Image Analysis(IRDE, DRDO)
PPT Image Analysis(IRDE, DRDO)PPT Image Analysis(IRDE, DRDO)
PPT Image Analysis(IRDE, DRDO)Nidhi Gopal
 
Feature Extraction
Feature ExtractionFeature Extraction
Feature Extractionskylian
 
Introduction to Image Processing with MATLAB
Introduction to Image Processing with MATLABIntroduction to Image Processing with MATLAB
Introduction to Image Processing with MATLABSriram Emarose
 

Viewers also liked (20)

Texture in image processing
Texture in image processing Texture in image processing
Texture in image processing
 
Image Texture Analysis
Image Texture AnalysisImage Texture Analysis
Image Texture Analysis
 
Features image processing and Extaction
Features image processing and ExtactionFeatures image processing and Extaction
Features image processing and Extaction
 
Matlab Feature Extraction Using Segmentation And Edge Detection
Matlab Feature Extraction Using Segmentation And Edge DetectionMatlab Feature Extraction Using Segmentation And Edge Detection
Matlab Feature Extraction Using Segmentation And Edge Detection
 
Improving Performance of Texture Based Face Recognition Systems by Segmenting...
Improving Performance of Texture Based Face Recognition Systems by Segmenting...Improving Performance of Texture Based Face Recognition Systems by Segmenting...
Improving Performance of Texture Based Face Recognition Systems by Segmenting...
 
Low level feature extraction - chapter 4
Low level feature extraction - chapter 4Low level feature extraction - chapter 4
Low level feature extraction - chapter 4
 
Analisis komponen utama (Principal Component Analysis)
Analisis komponen utama (Principal Component Analysis)Analisis komponen utama (Principal Component Analysis)
Analisis komponen utama (Principal Component Analysis)
 
Lbp based edge-texture features for object recoginition
Lbp based edge-texture features for object recoginitionLbp based edge-texture features for object recoginition
Lbp based edge-texture features for object recoginition
 
SVM-based CBIR of breast masses on mammograms
SVM-based CBIR of breast masses on mammogramsSVM-based CBIR of breast masses on mammograms
SVM-based CBIR of breast masses on mammograms
 
04 image enhancement in spatial domain DIP
04 image enhancement in spatial domain DIP04 image enhancement in spatial domain DIP
04 image enhancement in spatial domain DIP
 
Feature Extraction and Principal Component Analysis
Feature Extraction and Principal Component AnalysisFeature Extraction and Principal Component Analysis
Feature Extraction and Principal Component Analysis
 
Iris feature extraction
Iris feature extractionIris feature extraction
Iris feature extraction
 
BC Workflow
BC WorkflowBC Workflow
BC Workflow
 
2.6 support vector machines and associative classifiers revised
2.6 support vector machines and associative classifiers revised2.6 support vector machines and associative classifiers revised
2.6 support vector machines and associative classifiers revised
 
PPT Image Analysis(IRDE, DRDO)
PPT Image Analysis(IRDE, DRDO)PPT Image Analysis(IRDE, DRDO)
PPT Image Analysis(IRDE, DRDO)
 
Support Vector Machine
Support Vector MachineSupport Vector Machine
Support Vector Machine
 
Lecture12 - SVM
Lecture12 - SVMLecture12 - SVM
Lecture12 - SVM
 
End1
End1End1
End1
 
Feature Extraction
Feature ExtractionFeature Extraction
Feature Extraction
 
Introduction to Image Processing with MATLAB
Introduction to Image Processing with MATLABIntroduction to Image Processing with MATLAB
Introduction to Image Processing with MATLAB
 

Similar to Grey-level Co-occurence features for salt texture classification

Multiple UGV SLAM Map Sharing
Multiple UGV SLAM Map SharingMultiple UGV SLAM Map Sharing
Multiple UGV SLAM Map SharingAkash Borate
 
MrKNN_Soft Relevance for Multi-label Classification
MrKNN_Soft Relevance for Multi-label ClassificationMrKNN_Soft Relevance for Multi-label Classification
MrKNN_Soft Relevance for Multi-label ClassificationYI-JHEN LIN
 
Object detection at night
Object detection at nightObject detection at night
Object detection at nightSanjay Crúzé
 
various methods for image segmentation
various methods for image segmentationvarious methods for image segmentation
various methods for image segmentationRaveesh Methi
 
Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...
Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...
Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...Maninda Edirisooriya
 
presentation644v4
presentation644v4presentation644v4
presentation644v4Maikon
 
Detection&Tracking - Thermal imaging object detection and tracking
Detection&Tracking - Thermal imaging object detection and trackingDetection&Tracking - Thermal imaging object detection and tracking
Detection&Tracking - Thermal imaging object detection and trackinghadarpinhas1
 
Change detection techniques
Change detection techniquesChange detection techniques
Change detection techniquesFemi Opaleye
 
Temporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingTemporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingNAVER Engineering
 
Contour-Constrained Superpixels for Image and Video Processing
Contour-Constrained Superpixels for Image and Video ProcessingContour-Constrained Superpixels for Image and Video Processing
Contour-Constrained Superpixels for Image and Video ProcessingNAVER Engineering
 
Data mining techniques unit v
Data mining techniques unit vData mining techniques unit v
Data mining techniques unit vmalathieswaran29
 
CASTEP Software.pptx
CASTEP Software.pptxCASTEP Software.pptx
CASTEP Software.pptxMohsan10
 
Mean shift and Hierarchical clustering
Mean shift and Hierarchical clustering Mean shift and Hierarchical clustering
Mean shift and Hierarchical clustering Yan Xu
 
A Sparse-Coding Based Approach for Class-Specific Feature Selection
A Sparse-Coding Based Approach for Class-Specific Feature SelectionA Sparse-Coding Based Approach for Class-Specific Feature Selection
A Sparse-Coding Based Approach for Class-Specific Feature SelectionDavide Nardone
 

Similar to Grey-level Co-occurence features for salt texture classification (20)

Multiple UGV SLAM Map Sharing
Multiple UGV SLAM Map SharingMultiple UGV SLAM Map Sharing
Multiple UGV SLAM Map Sharing
 
MrKNN_Soft Relevance for Multi-label Classification
MrKNN_Soft Relevance for Multi-label ClassificationMrKNN_Soft Relevance for Multi-label Classification
MrKNN_Soft Relevance for Multi-label Classification
 
230727_HB_JointJournalClub.pptx
230727_HB_JointJournalClub.pptx230727_HB_JointJournalClub.pptx
230727_HB_JointJournalClub.pptx
 
Object detection at night
Object detection at nightObject detection at night
Object detection at night
 
various methods for image segmentation
various methods for image segmentationvarious methods for image segmentation
various methods for image segmentation
 
Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...
Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...
Lecture 11 - KNN and Clustering, a lecture in subject module Statistical & Ma...
 
presentation644v4
presentation644v4presentation644v4
presentation644v4
 
Detection&Tracking - Thermal imaging object detection and tracking
Detection&Tracking - Thermal imaging object detection and trackingDetection&Tracking - Thermal imaging object detection and tracking
Detection&Tracking - Thermal imaging object detection and tracking
 
Change detection techniques
Change detection techniquesChange detection techniques
Change detection techniques
 
Temporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingTemporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch Matching
 
Thesis presentation
Thesis presentationThesis presentation
Thesis presentation
 
face detection
face detectionface detection
face detection
 
Machine Learning
Machine LearningMachine Learning
Machine Learning
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
Contour-Constrained Superpixels for Image and Video Processing
Contour-Constrained Superpixels for Image and Video ProcessingContour-Constrained Superpixels for Image and Video Processing
Contour-Constrained Superpixels for Image and Video Processing
 
Data mining techniques unit v
Data mining techniques unit vData mining techniques unit v
Data mining techniques unit v
 
CASTEP Software.pptx
CASTEP Software.pptxCASTEP Software.pptx
CASTEP Software.pptx
 
Mean shift and Hierarchical clustering
Mean shift and Hierarchical clustering Mean shift and Hierarchical clustering
Mean shift and Hierarchical clustering
 
TransDreamer.pptx
TransDreamer.pptxTransDreamer.pptx
TransDreamer.pptx
 
A Sparse-Coding Based Approach for Class-Specific Feature Selection
A Sparse-Coding Based Approach for Class-Specific Feature SelectionA Sparse-Coding Based Approach for Class-Specific Feature Selection
A Sparse-Coding Based Approach for Class-Specific Feature Selection
 

Recently uploaded

Molecular markers- RFLP, RAPD, AFLP, SNP etc.
Molecular markers- RFLP, RAPD, AFLP, SNP etc.Molecular markers- RFLP, RAPD, AFLP, SNP etc.
Molecular markers- RFLP, RAPD, AFLP, SNP etc.Silpa
 
Gwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRL
Gwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRLGwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRL
Gwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRLkantirani197
 
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry Areesha Ahmad
 
development of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virusdevelopment of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virusNazaninKarimi6
 
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort ServiceCall Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort Serviceshivanisharma5244
 
300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptxryanrooker
 
Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.Silpa
 
Grade 7 - Lesson 1 - Microscope and Its Functions
Grade 7 - Lesson 1 - Microscope and Its FunctionsGrade 7 - Lesson 1 - Microscope and Its Functions
Grade 7 - Lesson 1 - Microscope and Its FunctionsOrtegaSyrineMay
 
Selaginella: features, morphology ,anatomy and reproduction.
Selaginella: features, morphology ,anatomy and reproduction.Selaginella: features, morphology ,anatomy and reproduction.
Selaginella: features, morphology ,anatomy and reproduction.Silpa
 
Thyroid Physiology_Dr.E. Muralinath_ Associate Professor
Thyroid Physiology_Dr.E. Muralinath_ Associate ProfessorThyroid Physiology_Dr.E. Muralinath_ Associate Professor
Thyroid Physiology_Dr.E. Muralinath_ Associate Professormuralinath2
 
Role of AI in seed science Predictive modelling and Beyond.pptx
Role of AI in seed science  Predictive modelling and  Beyond.pptxRole of AI in seed science  Predictive modelling and  Beyond.pptx
Role of AI in seed science Predictive modelling and Beyond.pptxArvind Kumar
 
Human genetics..........................pptx
Human genetics..........................pptxHuman genetics..........................pptx
Human genetics..........................pptxSilpa
 
GBSN - Microbiology (Unit 3)Defense Mechanism of the body
GBSN - Microbiology (Unit 3)Defense Mechanism of the body GBSN - Microbiology (Unit 3)Defense Mechanism of the body
GBSN - Microbiology (Unit 3)Defense Mechanism of the body Areesha Ahmad
 
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and SpectrometryFAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and SpectrometryAlex Henderson
 
Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.Silpa
 
Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.Silpa
 
Dr. E. Muralinath_ Blood indices_clinical aspects
Dr. E. Muralinath_ Blood indices_clinical  aspectsDr. E. Muralinath_ Blood indices_clinical  aspects
Dr. E. Muralinath_ Blood indices_clinical aspectsmuralinath2
 
Asymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 b
Asymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 bAsymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 b
Asymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 bSérgio Sacani
 
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRingsTransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRingsSérgio Sacani
 
Module for Grade 9 for Asynchronous/Distance learning
Module for Grade 9 for Asynchronous/Distance learningModule for Grade 9 for Asynchronous/Distance learning
Module for Grade 9 for Asynchronous/Distance learninglevieagacer
 

Recently uploaded (20)

Molecular markers- RFLP, RAPD, AFLP, SNP etc.
Molecular markers- RFLP, RAPD, AFLP, SNP etc.Molecular markers- RFLP, RAPD, AFLP, SNP etc.
Molecular markers- RFLP, RAPD, AFLP, SNP etc.
 
Gwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRL
Gwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRLGwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRL
Gwalior ❤CALL GIRL 84099*07087 ❤CALL GIRLS IN Gwalior ESCORT SERVICE❤CALL GIRL
 
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
 
development of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virusdevelopment of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virus
 
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort ServiceCall Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
 
300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx
 
Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.
 
Grade 7 - Lesson 1 - Microscope and Its Functions
Grade 7 - Lesson 1 - Microscope and Its FunctionsGrade 7 - Lesson 1 - Microscope and Its Functions
Grade 7 - Lesson 1 - Microscope and Its Functions
 
Selaginella: features, morphology ,anatomy and reproduction.
Selaginella: features, morphology ,anatomy and reproduction.Selaginella: features, morphology ,anatomy and reproduction.
Selaginella: features, morphology ,anatomy and reproduction.
 
Thyroid Physiology_Dr.E. Muralinath_ Associate Professor
Thyroid Physiology_Dr.E. Muralinath_ Associate ProfessorThyroid Physiology_Dr.E. Muralinath_ Associate Professor
Thyroid Physiology_Dr.E. Muralinath_ Associate Professor
 
Role of AI in seed science Predictive modelling and Beyond.pptx
Role of AI in seed science  Predictive modelling and  Beyond.pptxRole of AI in seed science  Predictive modelling and  Beyond.pptx
Role of AI in seed science Predictive modelling and Beyond.pptx
 
Human genetics..........................pptx
Human genetics..........................pptxHuman genetics..........................pptx
Human genetics..........................pptx
 
GBSN - Microbiology (Unit 3)Defense Mechanism of the body
GBSN - Microbiology (Unit 3)Defense Mechanism of the body GBSN - Microbiology (Unit 3)Defense Mechanism of the body
GBSN - Microbiology (Unit 3)Defense Mechanism of the body
 
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and SpectrometryFAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
 
Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.
 
Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.
 
Dr. E. Muralinath_ Blood indices_clinical aspects
Dr. E. Muralinath_ Blood indices_clinical  aspectsDr. E. Muralinath_ Blood indices_clinical  aspects
Dr. E. Muralinath_ Blood indices_clinical aspects
 
Asymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 b
Asymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 bAsymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 b
Asymmetry in the atmosphere of the ultra-hot Jupiter WASP-76 b
 
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRingsTransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
 
Module for Grade 9 for Asynchronous/Distance learning
Module for Grade 9 for Asynchronous/Distance learningModule for Grade 9 for Asynchronous/Distance learning
Module for Grade 9 for Asynchronous/Distance learning
 

Grey-level Co-occurence features for salt texture classification

  • 1. Grey-Level Co-occurrence features for salt texture classification Igor Orlov Master’s thesis presentation University of Oslo, Department of Informatics Supervisor: Anne H Schistad Solberg
  • 2. Content • Introduction • Problem statement and motivation • Background • Theoretical review of methods used in thesis • Implementation • Input data set, finding optimal GLCM parameters • Evaluation • Finding the best classifier, testing on training and test images • Summary
  • 3. Motivation • Salt structures are closely related to carbohydrates discovery • Segmentation of salt textures on seismic images is needed to build a correct velocity model • Time-consuming process, needs automation
  • 4. Problem statement Need to segment salt textures on seismic images using grey-level co- occurrence matrices (GLCMs) Two main goals: • Provide an algorithm and visualization tool for finding the optimal GLCM parameters, such as offset, direction, window size • Check whether distance matrix between centers of two classes can be used as a GLCM feature, if yes – estimate performance
  • 5. Seismic data acquisition • Initial velocity model is based on mapping of textures to known classes • Then updated iteratively • In areas of salt it is hard to build a velocity model, mapping is done manually or semi-aided on each step
  • 6. Cuts of marine seismic images Inline slice (vertical cut) Timeslice (horizontal cut) Data is stored in a 3D cube, can be cut in 2 directions:
  • 7. Salt texture mapping done manually
  • 8. Textures in seismic images - inline • Sub-horizontal reflectors (yellow) • Up-dipping and down-dipping reflectors (orange) • Salt structures (blue)
  • 9. Textures in seismic images - timeslice • Sub-horizontal reflectors (yellow) • Steep-dipping reflectors (orange) • Salt structures (blue) Not considered in this thesis
  • 10. Background for research methods • Grey-level co-occurrence matrices – GLCM • GLCM features • Distance metrics • Gaussian classifier
  • 11. Grey-level co-occurrence matrix • Texture description method – intensity change histogram • Matrix where element (x,y) represents a probability of changing a pixel intensity from x to y in a local window of size W when shifting in direction θ and offset d • Second order statistic • Used 5 directions – 0°,45°,90°,135° and isotropic
  • 12. GLCM features • Contrast – intensity contrast in a local window, 0 for const • Correlation • Energy – sum squared of GLCM elements • Homogeneity – closeness of distribution in diagonal • Entropy, Inertia, Cluster Shade etc.
  • 13. Distance metrics • Used to find a distance between class centers • Euclidean distance • Mahalanobis – takes into account in-class distribution (Σ – average covariance matrix), scale invariant
  • 14. Multivariate Gaussian classifier • Based on Bayes’ rule • Assumes Gaussian distribution • Very simple, trains fast, good enough • The likelihood that vector x belongs to class wj
  • 15. Part 2. Implementation and evaluation • Input dataset from North Sea • Finding optimal GLCM parameters, algorithm and visualization • Mahalanobis distance matrix • Training Gaussian classifier • Estimating classification results on training and test images
  • 16. Input image and mask Original image Mask. Class 1 – sedimentary rocks (sub- horizontal), class 2 – salt
  • 18. Image pre-processing • Input image 1401x1501, [-346;399] • Cut histogram tails, normalize and shift to [0;255] • No histogram equalization
  • 19. Finding the optimal parameters - algorithm • Choose direction, offset, window size • For each pixel of class 1, class 2 calculate the GLCM matrix • Matrix 16x16 can be presented as a feature vector 1x256 • Calculate the average GLCMs for two classes – class centers • Build a distance matrix where each element is a pixel-wise Mahalanobis distance between corresponding elements of class center matrices • Do the same for other direction, offset and window size, compare norm
  • 20. Mahalanobis distance calculation • Optimal parameters is where the norm of Mahalanobis distance matrix between centers of two classes is the greatest
  • 21. GLCM parameters • 16 grey levels, fixed. Thus, all GLCMs had size 16x16 • Window size 31x31, 51x51 • 5 directions • Which direction is expected to be best? • Offset 1..15 for 31x31, 1..20 for 51x51
  • 22. Visualization of distance matrices for 51x51 Horizontal Vertical • Displaying Mahalanobis distance between average GLCMs for two classes – salt and sedimentary rocks
  • 23. Visualization of distance matrices for 51x51 Diagonal Isotropic
  • 24. Norm of distance matrix, 31x31
  • 25. Norm of distance matrix, 51x51
  • 26. Distance matrix for optimal parameters • Suppress values around the center, enhance values in the middle and on 3-pixel distance around the center
  • 27. Own GLCM feature • Distance GLCM also shows which GLCM elements are more important for separation of two classes • Pixel-wise multiplication of GLCM for pixel with Mahalanobis distance matrix, then sum elements • Can also be used as a weighting window, to strengthen the GLCM elements which are more important for separation • Gives us a new set of features – weighted contrast (contrast of GLCM weighted with distMatrix), weighted energy etc.
  • 28. Feature list • For each pixel of class 1 and class 2 we have calculated the following ten features: • Variance • GLCM contrast + weighted • GLCM energy + weighted • GLCM correlation + weighted • GLCM homogeneity + weighted • Own GLCM feature – sum of weighted GLCM elements
  • 29. Gaussian classification • Trained classifier on one-, two-, three-feature subsets – taken manually in a random way • Some where correlated, overtraining • One feature is not enough • Tested on original image (can estimate absolute performance) and 2 test images – only visual estimation
  • 30. Single feature - Contrast
  • 31. Single feature – distance GLCM (own)
  • 32. Single feature – Energy (Weighted)
  • 33. Classification results for single feature • For training image, class mapping was provided • TN – true negative (correctly classified rocks), FP – false positive (rocks classified as salt), TP – true positive (correctly classified salt).
  • 34. Contrast and Correlation • Good TN and TP rate is not a guarantee of good results • TN – 94,17%, TP – 91,8%
  • 36. Energy, Own feature and Variance
  • 38. Contrast, Own and Weighted Energy
  • 39. Classification results for multiple features
  • 40. Test images from same dataset inlinekorr700 inlinekorr780
  • 41. Classification on test image inlinekorr700
  • 42. Classification on test image inlinekorr780
  • 43. Summary - Achieved results • Found the optimal GLCM parameters for separation of 2 classes: • Isotropic, offset 2, window 51x51 • Invented and tested own GLCM feature: • Mahalanobis distance between average GLCMs for two classes • Trained and tested a set of Gaussian classifiers to find the best feature combination for salt texture extraction • Visually best is Contrast, Own GLCM and Weighted Energy
  • 44. Summary – Disadvantages and drawbacks • Only dataset from North Sea was investigated • Maximum dimension of Gaussian classifiers was 3 • Only variance, standard Matlab GLCM features (4) and own GLCM feature were tested • Feature selection done manually • Only visual testing for test images
  • 45. Summary - Possible improvements • 3D GLCM – spatial relationship • Combination with other texture analysis methods – Fourier transform, greater order statistics etc. • Own feature – Mahalanobis distance - is based on pixel class mapping. Needs generalization (formula)
  • 46. The end Thanks for attention, questions…