Ter Haar Romeny, Computer Vision 2014 Geometry-driven diffusion: nonlinear scale-space – adaptive scale-space.

Slides:



Advertisements
Similar presentations
Johann Radon Institute for Computational and Applied Mathematics: 1/25 Signal- und Bildverarbeitung, Image Analysis and Processing.
Advertisements

Ter Haar Romeny, EMBS Berder 2004 Deblurring with a scale-space approach.
Ter Haar Romeny, FEV 2005 Curve Evolution Rein van den Boomgaard, Bart ter Haar Romeny Univ. van Amsterdam, Eindhoven University of technology.
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.
Scale & Affine Invariant Interest Point Detectors Mikolajczyk & Schmid presented by Dustin Lennon.
Various Regularization Methods in Computer Vision Min-Gyu Park Computer Vision Lab. School of Information and Communications GIST.
R. DOSIL, X. M. PARDO, A. MOSQUERA, D. CABELLO Grupo de Visión Artificial Departamento de Electrónica e Computación Universidade de Santiago de Compostela.
Active Contours, Level Sets, and Image Segmentation
Topic 6 - Image Filtering - I DIGITAL IMAGE PROCESSING Course 3624 Department of Physics and Astronomy Professor Bob Warwick.
Lecture 6 Sharpening Filters
Johann Radon Institute for Computational and Applied Mathematics: 1/33 Signal- und Bildverarbeitung, Image Analysis and Processing.
Johann Radon Institute for Computational and Applied Mathematics: 1/31 Signal- und Bildverarbeitung, Image Analysis and Processing.
Johann Radon Institute for Computational and Applied Mathematics: 1/39 The Gaussian Kernel, Regularization.
Signal- und Bildverarbeitung, 323
Johann Radon Institute for Computational and Applied Mathematics: 1/28 Signal- und Bildverarbeitung, Image Analysis and Processing.
Digital Image Processing
Image Segmentation some examples Zhiqiang wang
Edge and Corner Detection Reading: Chapter 8 (skip 8.1) Goal: Identify sudden changes (discontinuities) in an image This is where most shape information.
Edge detection Goal: Identify sudden changes (discontinuities) in an image Intuitively, most semantic and shape information from the image can be encoded.
Midterm Review CS485/685 Computer Vision Prof. Bebis.
1 Curvature Driven Flows Allen Tannenbaum. 2 Basic curve evolution: Invariant Flows  Planar curve:  General flow:  General geometric flow:
Edge detection. Edge Detection in Images Finding the contour of objects in a scene.
Image processing. Image operations Operations on an image –Linear filtering –Non-linear filtering –Transformations –Noise removal –Segmentation.
EE663 Image Processing Edge Detection 2 Dr. Samir H. Abdul-Jauwad Electrical Engineering Department King Fahd University of Petroleum & Minerals.
Computer Vision - A Modern Approach
Digital Image Processing
Diffusion Filters S. Derin Babacan Department of Electrical and Computer Engineering Northwestern University March 7, 2006 ECE 463 Term Project.
Linear filters and edges. Linear Filters General process: Form new image whose pixels are a weighted sum of original pixel values, using the same set.
CS443: Digital Imaging and Multimedia Filters Spring 2008 Ahmed Elgammal Dept. of Computer Science Rutgers University Spring 2008 Ahmed Elgammal Dept.
2-D, 2nd Order Derivatives for Image Enhancement
/mhj In Vivo Quantitative Assessment of Carotid Plaque component with multi-contrast MRI Jannie Wijnen 22 may 2003 using clustering algorithms, implemented.
Johann Radon Institute for Computational and Applied Mathematics: Summary and outlook.
Despeckle Filtering in Medical Ultrasound Imaging
Rician Noise Removal in Diffusion Tensor MRI
Johann Radon Institute for Computational and Applied Mathematics: 1/49 Signal- und Bildverarbeitung, silently converted to:
A Gentle Introduction to Bilateral Filtering and its Applications Limitation? Pierre Kornprobst (INRIA) 0:20.
Motivation from Real-World Applications EE565 Advanced Image Processing Copyright Xin Li Noisy Photos Noisy ultrasound data.
Machine Vision ENT 273 Image Filters Hema C.R. Lecture 5.
Comparison of Ventricular Geometry for Two Real-Time 3D Ultrasound Machines with Three-dimensional Level Set Elsa D. Angelini, Rio Otsuka, Shunishi Homma,
Top-points in Image Matching Bram Platel Evgenya Balmashnova Luc Florack Bart ter Haar Romeny.
Digital Image Processing Lecture 5: Neighborhood Processing: Spatial Filtering Prof. Charlene Tsai.
EDGE DETECTION IN COMPUTER VISION SYSTEMS PRESENTATION BY : ATUL CHOPRA JUNE EE-6358 COMPUTER VISION UNIVERSITY OF TEXAS AT ARLINGTON.
Edge Detection Today’s reading Cipolla & Gee on edge detection (available online)Cipolla & Gee on edge detection Szeliski, Ch 4.1.2, From Sandlot.
Edge detection Goal: Identify sudden changes (discontinuities) in an image Intuitively, most semantic and shape information from the image can be encoded.
CSE 185 Introduction to Computer Vision Edges. Scale space Reading: Chapter 3 of S.
Bart M. ter Haar Romeny.  Question: can top-points be used for object- retrieval tasks?
Ter Haar Romeny, CASIA 2015 scale-space The retina measures on many resolutions simultaneously.
Filtering Objective –improve SNR and CNR Challenges – blurs object boundaries and smears out important structures.
Scale-Space and Edge Detection Using Anisotropic Diffusion Presented By:Deepika Madupu Reference: Pietro Perona & Jitendra Malik.
Introduction to Scale Space and Deep Structure. Importance of Scale Painting by Dali Objects exist at certain ranges of scale. It is not known a priory.
Computer Vision Image Features Instructor: Dr. Sherif Sami Lecture 4.
Instructor: Mircea Nicolescu Lecture 7
Variational methods in image processing Variational methods in image processing Week 2: Diffusion Guy Gilboa Advanced Course
A Probabilistic Appraoch to Nonlinear Diffusion: Feature-Driven Filtering Hamid Krim ECE Dept. NCSU, Raleigh, NC
Miguel Tavares Coimbra
Digital Image Processing
Visual homing using PCA-SIFT
Edge Detection slides taken and adapted from public websites:
PDE Methods for Image Restoration
Prof. Bart M. ter Haar Romeny, PhD
ECE 692 – Advanced Topics in Computer Vision
Edge Detection CS 678 Spring 2018.
Lecture 3. Edge Detection, Texture
Edge detection Goal: Identify sudden changes (discontinuities) in an image Intuitively, most semantic and shape information from the image can be encoded.
ECE 692 – Advanced Topics in Computer Vision
Filtering Things to take away from this lecture An image as a function
Presented by Xu Miao April 20, 2005
Patricia van Marlen April 12, 2018
Presentation transcript:

ter Haar Romeny, Computer Vision 2014 Geometry-driven diffusion: nonlinear scale-space – adaptive scale-space

ter Haar Romeny, Computer Vision 2014 The advantage of selecting a larger scale is the improved reduction of noise, the appearance of more prominent structure, but the price to pay for this is reduced localization accuracy. Linear, isotropic diffusion cannot preserve the position of the differential invariant features over scale. A solution is to make the diffusion, i.e. the amount of blurring, locally adaptive to the structure of the image.

ter Haar Romeny, Computer Vision 2014 It takes geometric reasoning to come up with the right nonlinearity for the task, to include knowledge. We call this field evolutionary computing of image structure, or the application of evolutionary operations.

ter Haar Romeny, Computer Vision 2014 It is a divergence of a flow. We also call the flux function. With c = 1 we have normal linear, isotropic diffusion: the divergence of the gradient flow is the Laplacian. A conductivity coefficient (c) is introduced in the diffusion equation:

ter Haar Romeny, Computer Vision 2014

The Perona & Malik equation (1991): The conductivity terms are equivalent to first order. Paper P&M

ter Haar Romeny, Computer Vision 2014 The function k determines the weight of the gradient squared, i.e. how much is blurred at the edges. For the limit of k to infinity, we get linear diffusion again.

ter Haar Romeny, Computer Vision 2014 Working out the differentiations, we get a strongly nonlinear diffusion equation: The solution is not known analytically, so we have to rely on numerical methods, such as the forward Euler method: This process is an evolutionary computation. Nonlinear scale-space

ter Haar Romeny, Computer Vision 2014 Test on a small test image: Note the preserved steepness of the edges with the strongly reduced noise.

ter Haar Romeny, Computer Vision 2014 GDD is particularly useful for ultrasound edge-preserving speckle removal

ter Haar Romeny, Computer Vision 2014 Locally adaptive elongation of the diffusion kernel: Coherence Enhancing Diffusion Reducing to a small kernel is often not effective:

ter Haar Romeny, Computer Vision 2014 Coherence enhancing diffusion J. Weickert, 2001

ter Haar Romeny, Computer Vision 2014 Interestingly, the S/N ratio increases during the evolution. Signal = mean difference Noise = variance SNR = S/N

ter Haar Romeny, Computer Vision 2014 Interestingly, the S/N ratio has a maximum during the evolution. Blurring too long may lead to complete blurring (always some leak). Signal = mean difference Noise = variance SNR = S/N

ter Haar Romeny, Computer Vision 2014 The Perona & Malik equation is ill-posed.

ter Haar Romeny, Computer Vision 2014

Atherosclerotic plaque classification – P&M, k = 50,  s = 5 J. Wijnen, TUE-BME, 2003

ter Haar Romeny, Computer Vision 2014

Stability of the numerical implementation The maximal step size is limited due to the Neumann stability criterion (1/4). The maximal step size when using Gaussian derivatives is substantially larger:

ter Haar Romeny, Computer Vision 2014 Test image and its blurred version. Note the narrow range of the allowable time step maximum.

ter Haar Romeny, Computer Vision 2014 Euclidean shortening flow: Working out the derivatives, this is L vv in gauge coordinates, i.e. the ridge detector. So: Because the Laplacian is we get We see that we have corrected the normal diffusion with a factor proportional to the second order derivative in the gradient direction (in gauge coordinates: ). This subtractive term cancels the diffusion in the direction of the gradient.

ter Haar Romeny, Computer Vision 2014 This diffusion scheme is called Euclidean shortening flow due to the shortening of the isophotes, when considered as a curve-evolution scheme. Advantage: there is no parameter k. Disadvantage: rounding of corners.

ter Haar Romeny, Computer Vision 2014