Perceptual grouping: Curvature enhanced closure of elongated structures By Gijs Huisman Committee: prof. dr. ir. B.M. ter Haar Romeny prof. dr. ir. P.

Slides:



Advertisements
Similar presentations
SPM – introduction & orientation introduction to the SPM software and resources introduction to the SPM software and resources.
Advertisements

Basis beeldverwerking (8D040) dr. Andrea Fuster dr. Anna Vilanova Prof.dr.ir. Marcel Breeuwer Filtering.
Ter Haar Romeny, FEV MIT AI Lab Automatic Polyp Detection.
Ter Haar Romeny, TU/e Mathematical Models of Contextual Operators Eindhoven University of Technology Department of Biomedical Engineering Markus van Almsick,
Ter Haar Romeny, EMBS Berder 2004 Gaussian derivative kernels The algebraic expressions for the first 4 orders:
Ter Haar Romeny, FEV Geometry-driven diffusion: nonlinear scale-space – adaptive scale-space.
Ter Haar Romeny, ICPR 2010 Introduction to Scale-Space and Deep Structure.
TU/e, BMIA & PMS, X-Ray Predevelopment, Erik Franken, Context-Enhanced Detection of Electrophysiology Catheters in Noisy Fluoroscopy Images.
Probabilistic Inverse Dynamics for Nonlinear Blood Pattern Reconstruction Benjamin Cecchetto, Wolfgang Heidrich University of British Columbia.
Developable Surface Fitting to Point Clouds Martin Peternell Computer Aided Geometric Design 21(2004) Reporter: Xingwang Zhang June 19, 2005.
R. DOSIL, X. M. PARDO, A. MOSQUERA, D. CABELLO Grupo de Visión Artificial Departamento de Electrónica e Computación Universidade de Santiago de Compostela.
Introduction to the Curvelet Transform
Spatial Filtering (Chapter 3)
Ter Haar Romeny, Computer Vision 2014 Geometry-driven diffusion: nonlinear scale-space – adaptive scale-space.
Johann Radon Institute for Computational and Applied Mathematics: 1/39 The Gaussian Kernel, Regularization.
An Adaptive Image Enhancement Algorithm for Face Detection By Lizuo Jin, Shin’ichi Satoh, and Masao Sakauchi. ECE 738 In Young Chung.
Digital Image Processing
6/9/2015Digital Image Processing1. 2 Example Histogram.
Surface Reconstruction from 3D Volume Data. Problem Definition Construct polyhedral surfaces from regularly-sampled 3D digital volumes.
Real-time Embedded Face Recognition for Smart Home Fei Zuo, Student Member, IEEE, Peter H. N. de With, Senior Member, IEEE.
Image Filtering CS485/685 Computer Vision Prof. George Bebis.
Digital Image Processing: Revision
Challenges. Challenge I : Improve adaptive diffusion on orientation score Iterate: 1. Estimate local shape (by exp-curve fit ) 2. Diffusion-step curvature.
Enhancement, Completion and Detection of Elongated Structures in Medical Imaging via Evolutions on Lie Groups muscle cells bone-structure retinal bloodvessels.
Perceptual Organization A Mathematical Approach Based on scale, orientation and curvature Remco Duits.
1 1 Contour Enhancement and Completion via Left-Invariant Second Order Stochastic Evolution Equations on the 2D-Euclidean Motion Group Erik Franken, Remco.
Perceptual Organization A Mathematical Approach Based on scale and orientation Remco Duits.
Am Analysis of Coronary Microvessel Structures on the Enhancement and Detection of Microvessels in 3D Cryomicrotome Data Master’s project by Edwin Bennink.
Tools for Shape Analysis of Vascular Response using Two Photon Laser Scanning Microscopy By Han van Triest Committee: Prof. Dr. Ir. B.M. ter Haar Romeny.
Perceptual Grouping: The Closure of Gaps within Elongated Structures in Medical Images Renske de Boer March 23 rd, /mhj Committee: prof. dr.
An Overview of Cores Yoni Fridman The University of North Carolina at Chapel Hill Medical Image Display & Analysis Group Based on work by Fridman, Furst,
Computational Photography: Image Processing Jinxiang Chai.
The automation of generalized curves method presentation on the map at any scales Prof. Tadeusz Chrobak AGH University of Science and Technology Poland.
Gwangju Institute of Science and Technology Intelligent Design and Graphics Laboratory Multi-scale tensor voting for feature extraction from unstructured.
Spatial Filtering: Basics
October 14, 2014Computer Vision Lecture 11: Image Segmentation I 1Contours How should we represent contours? A good contour representation should meet.
1 Chapter 5 Image Transforms. 2 Image Processing for Pattern Recognition Feature Extraction Acquisition Preprocessing Classification Post Processing Scaling.
Basis beeldverwerking (8D040) dr. Andrea Fuster Prof.dr. Bart ter Haar Romeny dr. Anna Vilanova Prof.dr.ir. Marcel Breeuwer Convolution.
University of Texas at Austin CS384G - Computer Graphics Fall 2010 Don Fussell Image processing.
Automatic Minirhizotron Root Image Analysis Using Two-Dimensional Matched Filtering and Local Entropy Thresholding Presented by Guang Zeng.
Detection of nerves in Ultrasound Images using edge detection techniques NIRANJAN TALLAPALLY.
Basis Expansions and Regularization Part II. Outline Review of Splines Wavelet Smoothing Reproducing Kernel Hilbert Spaces.
Lecture 5 Mask/Filter Transformation 1.The concept of mask/filters 2.Mathematical model of filtering Correlation, convolution 3.Smoother filters 4.Filter.
Image Subtraction Mask mode radiography h(x,y) is the mask.
Course 2 Image Filtering. Image filtering is often required prior any other vision processes to remove image noise, overcome image corruption and change.
MOS Data Reduction Michael Balogh University of Durham.
Mestrado em Ciência de Computadores Mestrado Integrado em Engenharia de Redes e Sistemas Informáticos VC 15/16 – TP7 Spatial Filters Miguel Tavares Coimbra.
Non-linear Filters Non-linear filter (nelineární filtr) –spatial non-linear operator that produces the output image array g(x,y) from the input image array.
İmage enhancement Prepare image for further processing steps for specific applications.
Image enhancement Last update Heejune Ahn, SeoulTech.
Ship Computer Aided Design
Instructor: Mircea Nicolescu Lecture 5 CS 485 / 685 Computer Vision.
Digital Image Processing Lecture 9: Filtering in Frequency Domain Prof. Charlene Tsai.
Frank Bergschneider February 21, 2014 Presented to National Instruments.
Detection of nerves in Ultrasound Images using edge detection techniques NIRANJAN TALLAPALLY.
Filters– Chapter 6. Filter Difference between a Filter and a Point Operation is that a Filter utilizes a neighborhood of pixels from the input image to.
Medical Image Analysis
Miguel Tavares Coimbra
Image Subtraction Mask mode radiography h(x,y) is the mask.
Invertible Orientation Scores of 3D Images
Edge Detection slides taken and adapted from public websites:
Basis beeldverwerking (8D040) dr. Andrea Fuster dr. Anna Vilanova Prof
Lecture 3 (2.5.07) Image Enhancement in Spatial Domain
CSC 381/481 Quarter: Fall 03/04 Daniela Stan Raicu
Pattern Recognition and Machine Learning
Region and Shape Extraction
Digital Image Processing Lecture 11: Image Restoration
Image Enhancement in the Spatial Domain
Introduction to Artificial Intelligence Lecture 22: Computer Vision II
Goodfellow: Chapter 14 Autoencoders
Presentation transcript:

Perceptual grouping: Curvature enhanced closure of elongated structures By Gijs Huisman Committee: prof. dr. ir. B.M. ter Haar Romeny prof. dr. ir. P. Hilbers dr. L.M.J. Florack dr. ir. R. Duits ir. E.M. Franken

2/30 Content 1.Introduction 2.Orientation scores Cake kernels 3.G-convolution Stochastic completion kernel Adaptive G-Convolution 4.Mode line extraction Theory 5.Non-linear operations Advection based enhancement 3 non-linear operations 6.Curvature estimation 4 methods Test results 7.Experiments Mode line extraction Increased gap filling Improved smoothness Adaptive shooting Examples medical images 8.Conclusion Conclusions Recommendations

3/30 Introduction

4/30 Orientation score An orientation score has 2 spatial dimensions and 1 orientation dimension

5/30 Orientation Score An orientation score is obtained by wavelet transformation of an image Where and Reconstruction of an image is possible by an inverse wavelet transform

6/30 Orientation score Cake Kernels The waveletis defined by: The function is defined by B- splines: Main advantage is easily adaptive kernels with good reconstruction properties is defined by a 2D gauss

7/30 G-convolution Normal convolution G-convolution

8/30 Stochastic Completion Kernel G-convolution The used kernel depicts a probability density function for the continuation of a line kernel in an orientation score.

9/30 G-convolution Stochastic Completion Kernel Gap closing operation with the stochastic completion kernel

10/30 G-convolution Making the G-convolution adaptive means that the kernel properties change with the position in the orientation score. Kernels are adapted to fit the local curvature Adaptive

11/30 Mode line extraction Very often the lines itself are demanded instead of an enhanced image. Any point is part of a local mode line if and at the point Lines in the spatial plane are 3D ridges in an orientation score.

12/30 Non-Linear Operations Enhancement can be done before and after an G-convolution Non ideal cake kernel response: DC-extraction MIN-Extraction Erosion Advection

13/30 Non-Linear Operations DC-Extraction MIN-Extraction Erosion

14/30 Non-Linear Operations Advectio n A force field directed towards the local mode lines: By means of advection the score can now be sharpened

15/30 Non-Linear Operations Results Erosion DC-extractionMIN-extraction Straight Curved Advection Intensity No preprocessing

16/30 Curvature estimation 1.Inner product stochastic completion kernel 2.Inner product Gaussian based kernel 3.Region estimation 4.Hessian based method

17/30 Curvature estimation Results Stochastic Gaussian Region Hessian

18/30 Curvature estimation Results Curvature measurement on a cross section of the circle line Stochastic Gaussian Regio n Hessian

19/30 Curvature estimation Results Stochastic Gaussian Region Hessian

20/30 Experiments Mode line extraction

21/30 Experiments Mode line extraction

22/30 Experiments Mode line extraction Mode line extraction on artificial image

23/30 Experiments Increased gap filling Plane DCMin Plane DCMin

24/30 Experiments Improved smoothness Straight Curved

25/30 Experiments Adaptive shooting Original imageOrientation score Straight shooting result Curvature estimate Enhanced image

26/30 Experiments Adaptive shooting OriginalStraight shooting (1)Curved Shooting (2)Curved Shooting (3) Mean Filling Method Min Filling Method

27/30 Experiments Examples Medical images Blood vessel extraction on images of the human retina Original Threshold Straight shooting Blood vessels

28/30 Experiments Examples Medical images Straight shooting Adaptive shooting

29/30 Conclusion Curvature enhanced shooting does improve the gap filling Successful method of curve extraction Good method to estimate the curvature Improve the accuracy of the curve extraction method Better numerical implementation advection enhancement Devise a method to extract the correct curves (e.g. fast marching) Better tuning of the cake kernel parameters Conclusions Recommendations

30/30 Questions?