CN780 Final Lecture. Low-Level Vision, Scale-Space, and Polyakov Action. Neil I. Weisenfeld

Size: px
Start display at page:

Download "CN780 Final Lecture. Low-Level Vision, Scale-Space, and Polyakov Action. Neil I. Weisenfeld"

Transcription

1 CN780 Final Lecture Low-Level Vision, Scale-Space, and Polyakov Action Neil I. Weisenfeld Department of Cognitive and Neural Systems Boston University chapter May 9, 2005 p.1/25

2 Introduction The term early vision refers to the processing that happens, or is thought to happen, at the beginning of the visual analysis chain close to the retinal image. At least since Hubel and Wiesel s discovery of orientation sensitive cells in V1, scientists have been obsessed with models of early vision that center around the extraction of features, especially edges, from a scene. Computational strategies focus on context-free and therefore local image-processing operations. As Sochen points out, this locality leads to differential operators. chapter May 9, 2005 p.2/25

3 Laplacian Edge Detection Image on the left is the original, blurred image. Image in the middle is the log-magnitude of the Laplacian applied to the image. Image on the right is the zero-crossing of the Laplacian image. elec539/projects97/morphj chapter May 9, 2005 p.3/25

4 Scale-Space: Isotropic Diffusion Some authors propose a pyramid where images are smooth at various resolutions. But this can be generalized to a continuum: I(x, y, t) = I 0 (x, y) G(x, y; t) Perona-Malik note that this is also the solution to the heat equation and therefore has a physical interpretation: isotropic diffusion: I t = I = (I xx + I yy ) chapter May 9, 2005 p.4/25

5 Scale-Space: Isotropic Diffusion But isotropic diffusion has two problems: edges are not preserved and they move at different scales: chapter May 9, 2005 p.5/25

6 Scale-Space: Anisotropic Diffusion Perona and Malik introduced the idea of anisotropic diffusion where the constant in the heat equation is now variable with space. I t = div(c(x, y, t) I) = c(x, y, t) I + c I if c(x, y, t) = c then we have I t = c I now choose c in a way that takes the gradient into account: c(x, y, t) = g( I(x, y, t) ) chapter May 9, 2005 p.6/25

7 Scale-Space: Anisotropic vs. Isotropic Diffusion chapter May 9, 2005 p.7/25

8 Sochen Unifying Theory Sochen, 1996, From High Energy Physics to Low Level Vision (Lawrence Berkeley National Labs Tech Report LBNL-39243) Images modeled as a surface embedded in a higher dimensional space (e.g. a 2D image as a height-field in R 3 ). Image processing operations that generate a scale space are modeled as a metric in the embedded space and a flow. The Polyakov action, from high-energy physics, is used to generate the flow. A number of classic operations fall into this framework and the mathematics are in place for extension to other problems. Sochen, for example, shows how to use the theory to extend some classical operations on grayscale images to RGB-color images. chapter May 9, 2005 p.8/25

9 Image as a Surface 2D, single-channel image I(x, y) to R 3 as (x, y, I) I(x, y) (x, y, I) chapter May 9, 2005 p.9/25

10 Where are we going with this? We re interested in differential operations on, say, image intensity values with respect to image coordinates. The surface embedding allows us to visualize this in two ways, either in (σ 1, σ 2 ) space (with a local metric) or (x, y, I): chapter May 9, 2005 p.10/25

11 Differential Geometry Intro How do we measure distances between points within our coordinate systems? Consider the map X : Σ R 3 X is explicitly [X 1 (σ 1, σ 2 ), X 2 (σ 1, σ 2 ), X 3 (σ 1, σ 2 )] Squared distance ds 2 on Σ is local coordinates adjusted by g µν (σ 1, σ 2 ): ds 2 = g µν dσ µ dσ ν Einstein summation = g 11 (dσ 1 ) 2 + 2g 12 dσ 1 dσ 2 + g 22 (dσ 2 ) 2 chapter May 9, 2005 p.11/25

12 Connect Metrics Consider a more general map X : Σ M and metric g on Σ and metric h on M. If we know h and X, we can construct g: g µν (σ 1, σ 2 ) = h ij (X) µ X i ν X j note we re summing over i, j and µ X i means X i / σ µ For our image map: X : (σ 1, σ 2 ) [x = σ 1, y = σ 2, z = I(σ 1, σ 2 )] we have: ( ) 1 + Ix 2 I x I y (g µν ) = I x I y 1 + Iy 2 chapter May 9, 2005 p.12/25

13 Where are we? In the grand tradition, and hopefully more successfully that Sun Microsystems campaign The Network is the Computer : note that The Map is the Image. Or, at least, it contains the image. We want to (remember) define a scale-space and flows (transitions) from one image to another. The framework calls for a measure on the maps and equations will be derived by minimizing the measure with respect to something. chapter May 9, 2005 p.13/25

14 Polyakov Action: The Measure Again, we have (Σ, g) and (M, h) and our mapping X : Σ M A measure of the weight of the map is: S[X i, g µν, h ij ] = d m σ gg µν µ X i ν X j h ij (X) m = dimension(σ), g = g µν, g µν = g 1 µν rememember that g µν = µ X i ν X j h ij (X) and g µν g νγ = δ µ γ, so the weight is related to the surface area. This functional is thought to have been introduced by Polyakov for m = 2. chapter May 9, 2005 p.14/25

15 Simple Example Embed a surface in R 3 and use Cartesian coordinates, so: ( ) 1 0 (g µν ) =, x = σ 1, y = σ 2, h ij = δ ij 0 1 The functional is now the Euclidean L2 norm: S[I, g µν = δ µν, h ij = δ ij ] = d 2 σ( x 2 + y 2 + I 2 ) and minimizing with respect to I gives the heat operator (c.f. isotropic diffusion, but we don t have a time factor, yet. Part of the flexibility here is deciding what to minimize with respect to and scaling of intensity via h ij chapter May 9, 2005 p.15/25

16 Minimizing a Functional The Euler-Lagrange equations can be used to find an extremum. Given: I = x2 f[y(x), ẏ(x), x]dx x1 Euler-Lagrange gets you: f y d dx ( f ẏ ) = 0 chapter May 9, 2005 p.16/25

17 Minimizing and Adding Time The details of how to do this are worked out in Appendix A of the difficult to find LBNL technical report and are eliminated from all of the published works. Minimizing the Polyakov functional with respect to the embedding space yields: 1 2 g δs hil =... (see paper) δxl Add time (scale-space) by viewing this as a gradient descent problem: X i = 1 t 2 δs hil g δx l chapter May 9, 2005 p.17/25

18 Perona-Malik Flow Perona-Malik Diffusion can be derived within this framework and has the advantage of extending to n-dimensions. The Euler-Lagrange equation simplifies, in a Euclidean space to: I t = µ gg µν ν I with the n-dimensional image embedded in R n+1 if we choose (g µν ) = fi d where I d is the identity, we have: I t = n µ=1 µ f n 2 1 µ I chapter May 9, 2005 p.18/25

19 Perona-Malik Flow cnt d For n 2, we can choose f n 2 1 = C(I) and we have Perona-Malik Flow: for C(I) = f(i 0) I, we have: I t = div(c(i) I) I t = div(f(i 0 ) I I ) which is the fundamental equation for geodesic active contours (c.f. level set methods) chapter May 9, 2005 p.19/25

20 Beltrami Flow The authors propose a new flow, operating in multiple dimensions, which the dub the Beltrami flow: I i t = g I g µ ( gg µν ν I i ) = H i where H i is the mean curvature in the ith direction. Geometrically, each point moves with a velocity proportional to the mean curvature in each direction. This looks nasty, but it is the induced flow when using the trivial embedding, Euclidean metrics, and minimizing with respect to I. See examples next page. chapter May 9, 2005 p.20/25

21 Beltrami Flow Examples chapter May 9, 2005 p.21/25

22 Multiscale Active Contours Bresson, et. al. used the framework to extend active contours to multiple scales: chapter May 9, 2005 p.22/25

23 Final Words on the Framework The framework provides: an embedding of images in a higher dimensional space via a mapping a connection between the metric in the image manifold and the (n + 1)D space an action functional on the maps themselves (which encode the images) The connection to established methods seems somewhat forced I need to understand the geometry better The technique opens up avenues for analysis and derivation: metrics can be chosen in either space and derived for the other and the action can be minimized with respect to the metrics or the mapping. chapter May 9, 2005 p.23/25

24 Concluding Ideas The authors cite 1994 work of Florac, et. al., in which they present linear heat flow scale-space in the log-polar coordinate system. They point out that the Beltrami operator is a parameterization invariant differential operator which can work in this capacity and is edge preserving. The generalization of Perona-Malik flow to n-dimensions may be of interest in multi-channel MRI. I have never seen precisely the same text recycled into so many different publications. Many of the details are worked out only in the LBNL tech report. The generalization of various techniques to n-dimensions may be of great interest in and of itself. chapter May 9, 2005 p.24/25

25 References Primary reference is Sochen, et. al.: A General Framework for Low Level Vision. IEEE Trans Image Proc 7(3): , The LBNL tech report is not available online from LBNL and the version of Sochen s web site is stripped down for one of the many re-publications of the text. The full version was on Kimmel s website at: ron/pub.html Sochen has a talk at sochen/beltrami2.pdf Multiscale Active Contours was from Bresson, et al. at EPFL, Lausanne, Switzerland. It does not appear to have been published yet. Malik and Perona is in IEEE PAMI 12(7): , chapter May 9, 2005 p.25/25

Efficient Beltrami Filtering of Color Images via Vector Extrapolation

Efficient Beltrami Filtering of Color Images via Vector Extrapolation Efficient Beltrami Filtering of Color Images via Vector Extrapolation Lorina Dascal, Guy Rosman, and Ron Kimmel Computer Science Department, Technion, Institute of Technology, Haifa 32000, Israel Abstract.

More information

R. Kimmel, 1 R. Malladi, 1 and N. Sochen 2. Lawrence Berkeley National Laboratory University of California, Berkeley,

R. Kimmel, 1 R. Malladi, 1 and N. Sochen 2. Lawrence Berkeley National Laboratory University of California, Berkeley, Image Processing via the Beltrami Operator? R. Kimmel, 1 R. Malladi, 1 and N. Sochen 2 1 Lawrence Berkeley National Laboratory University of California, Berkeley, CA 94720. 2 Electrical Engineering Dept.

More information

A Riemannian Framework for Denoising Diffusion Tensor Images

A Riemannian Framework for Denoising Diffusion Tensor Images A Riemannian Framework for Denoising Diffusion Tensor Images Manasi Datar No Institute Given Abstract. Diffusion Tensor Imaging (DTI) is a relatively new imaging modality that has been extensively used

More information

IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. XX, NO. Y, MONTH A General Framework for Low Level Vision. Nir Sochen Ron Kimmel Ravi Malladi

IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. XX, NO. Y, MONTH A General Framework for Low Level Vision. Nir Sochen Ron Kimmel Ravi Malladi IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. XX, NO. Y, MONTH 1999 100 A General Framework for Low Level Vision Nir Sochen Ron Kimmel Ravi Malladi Abstract We introduce a new geometrical framework based

More information

Where is the bark, the tree and the forest, mathematical foundations of multi-scale analysis

Where is the bark, the tree and the forest, mathematical foundations of multi-scale analysis Where is the bark, the tree and the forest, mathematical foundations of multi-scale analysis What we observe is not nature itself, but nature exposed to our method of questioning. -Werner Heisenberg M.

More information

On Mean Curvature Diusion in Nonlinear Image Filtering. Adel I. El-Fallah and Gary E. Ford. University of California, Davis. Davis, CA

On Mean Curvature Diusion in Nonlinear Image Filtering. Adel I. El-Fallah and Gary E. Ford. University of California, Davis. Davis, CA On Mean Curvature Diusion in Nonlinear Image Filtering Adel I. El-Fallah and Gary E. Ford CIPIC, Center for Image Processing and Integrated Computing University of California, Davis Davis, CA 95616 Abstract

More information

Introduction to Nonlinear Image Processing

Introduction to Nonlinear Image Processing Introduction to Nonlinear Image Processing 1 IPAM Summer School on Computer Vision July 22, 2013 Iasonas Kokkinos Center for Visual Computing Ecole Centrale Paris / INRIA Saclay Mean and median 2 Observations

More information

Medical Image Analysis

Medical Image Analysis Medical Image Analysis CS 593 / 791 Computer Science and Electrical Engineering Dept. West Virginia University 23rd January 2006 Outline 1 Recap 2 Edge Enhancement 3 Experimental Results 4 The rest of

More information

Tensor-based Image Diffusions Derived from Generalizations of the Total Variation and Beltrami Functionals

Tensor-based Image Diffusions Derived from Generalizations of the Total Variation and Beltrami Functionals Generalizations of the Total Variation and Beltrami Functionals 29 September 2010 International Conference on Image Processing 2010, Hong Kong Anastasios Roussos and Petros Maragos Computer Vision, Speech

More information

Nonlinear Diffusion. 1 Introduction: Motivation for non-standard diffusion

Nonlinear Diffusion. 1 Introduction: Motivation for non-standard diffusion Nonlinear Diffusion These notes summarize the way I present this material, for my benefit. But everything in here is said in more detail, and better, in Weickert s paper. 1 Introduction: Motivation for

More information

The Laplace-Beltrami operator: a ubiquitous tool for image and shape processing

The Laplace-Beltrami operator: a ubiquitous tool for image and shape processing The Laplace-Beltrami operator: a ubiquitous tool for image and shape processing Aaron Wetzler, Yonathan Aflalo, Anastasia Dubrovina, and Ron Kimmel Technion - Israel Institute of Technology, Haifa, Israel

More information

Lecture: Lorentz Invariant Dynamics

Lecture: Lorentz Invariant Dynamics Chapter 5 Lecture: Lorentz Invariant Dynamics In the preceding chapter we introduced the Minkowski metric and covariance with respect to Lorentz transformations between inertial systems. This was shown

More information

ON THE DISCRETE MAXIMUM PRINCIPLE FOR THE BELTRAMI COLOR FLOW

ON THE DISCRETE MAXIMUM PRINCIPLE FOR THE BELTRAMI COLOR FLOW ON THE DISCRETE MAXIMUM PRINCIPLE FOR THE BELTRAMI COLOR FLOW LORINA DASCAL, ADI DITKOWSKI, AND NIR A. SOCHEN Abstract. We analyze the discrete maximum principle for the Beltrami color flow. The Beltrami

More information

Nonlinear Diffusion. Journal Club Presentation. Xiaowei Zhou

Nonlinear Diffusion. Journal Club Presentation. Xiaowei Zhou 1 / 41 Journal Club Presentation Xiaowei Zhou Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology 2009-12-11 2 / 41 Outline 1 Motivation Diffusion process

More information

Courbure discrète : théorie et applications

Courbure discrète : théorie et applications Courbure discrète : théorie et applications Rencontre organisée par : Laurent Najman and Pascal Romon 18-22 novembre 2013 Yonathan Aflalo, Anastasia Dubrovina, Ron Kimmel, and Aaron Wetzler Curvature in

More information

Edge Detection. CS 650: Computer Vision

Edge Detection. CS 650: Computer Vision CS 650: Computer Vision Edges and Gradients Edge: local indication of an object transition Edge detection: local operators that find edges (usually involves convolution) Local intensity transitions are

More information

A Tutorial on Active Contours

A Tutorial on Active Contours Elham Sakhaee March 4, 14 1 Introduction I prepared this technical report as part of my preparation for Computer Vision PhD qualifying exam. Here we discuss derivation of curve evolution function for some

More information

Linear Diffusion. E9 242 STIP- R. Venkatesh Babu IISc

Linear Diffusion. E9 242 STIP- R. Venkatesh Babu IISc Linear Diffusion Derivation of Heat equation Consider a 2D hot plate with Initial temperature profile I 0 (x, y) Uniform (isotropic) conduction coefficient c Unit thickness (along z) Problem: What is temperature

More information

Vector and Tensor Calculus

Vector and Tensor Calculus Appendices 58 A Vector and Tensor Calculus In relativistic theory one often encounters vector and tensor expressions in both three- and four-dimensional form. The most important of these expressions are

More information

ITK Filters. Thresholding Edge Detection Gradients Second Order Derivatives Neighborhood Filters Smoothing Filters Distance Map Image Transforms

ITK Filters. Thresholding Edge Detection Gradients Second Order Derivatives Neighborhood Filters Smoothing Filters Distance Map Image Transforms ITK Filters Thresholding Edge Detection Gradients Second Order Derivatives Neighborhood Filters Smoothing Filters Distance Map Image Transforms ITCS 6010:Biomedical Imaging and Visualization 1 ITK Filters:

More information

C2 Differential Equations : Computational Modeling and Simulation Instructor: Linwei Wang

C2 Differential Equations : Computational Modeling and Simulation Instructor: Linwei Wang C2 Differential Equations 4040-849-03: Computational Modeling and Simulation Instructor: Linwei Wang Part II Variational Principle Calculus Revisited Partial Derivatives Function of one variable df dx

More information

Linear Diffusion and Image Processing. Outline

Linear Diffusion and Image Processing. Outline Outline Linear Diffusion and Image Processing Fourier Transform Convolution Image Restoration: Linear Filtering Diffusion Processes for Noise Filtering linear scale space theory Gauss-Laplace pyramid for

More information

NON-LINEAR DIFFUSION FILTERING

NON-LINEAR DIFFUSION FILTERING NON-LINEAR DIFFUSION FILTERING Chalmers University of Technology Page 1 Summary Introduction Linear vs Nonlinear Diffusion Non-Linear Diffusion Theory Applications Implementation References Page 2 Introduction

More information

General Relativity I

General Relativity I General Relativity I presented by John T. Whelan The University of Texas at Brownsville whelan@phys.utb.edu LIGO Livingston SURF Lecture 2002 July 5 General Relativity Lectures I. Today (JTW): Special

More information

Fraunhofer Institute for Computer Graphics Research Interactive Graphics Systems Group, TU Darmstadt Fraunhoferstrasse 5, Darmstadt, Germany

Fraunhofer Institute for Computer Graphics Research Interactive Graphics Systems Group, TU Darmstadt Fraunhoferstrasse 5, Darmstadt, Germany Scale Space and PDE methods in image analysis and processing Arjan Kuijper Fraunhofer Institute for Computer Graphics Research Interactive Graphics Systems Group, TU Darmstadt Fraunhoferstrasse 5, 64283

More information

Laplacian Mesh Processing

Laplacian Mesh Processing Sorkine et al. Laplacian Mesh Processing (includes material from Olga Sorkine, Yaron Lipman, Marc Pauly, Adrien Treuille, Marc Alexa and Daniel Cohen-Or) Siddhartha Chaudhuri http://www.cse.iitb.ac.in/~cs749

More information

Medical Image Analysis

Medical Image Analysis Medical Image Analysis CS 593 / 791 Computer Science and Electrical Engineering Dept. West Virginia University 20th January 2006 Outline 1 Discretizing the heat equation 2 Outline 1 Discretizing the heat

More information

Solving the Geodesic Equation

Solving the Geodesic Equation Solving the Geodesic Equation Jeremy Atkins December 12, 2018 Abstract We find the general form of the geodesic equation and discuss the closed form relation to find Christoffel symbols. We then show how

More information

Image enhancement. Why image enhancement? Why image enhancement? Why image enhancement? Example of artifacts caused by image encoding

Image enhancement. Why image enhancement? Why image enhancement? Why image enhancement? Example of artifacts caused by image encoding 13 Why image enhancement? Image enhancement Example of artifacts caused by image encoding Computer Vision, Lecture 14 Michael Felsberg Computer Vision Laboratory Department of Electrical Engineering 12

More information

Face Recognition Using Laplacianfaces He et al. (IEEE Trans PAMI, 2005) presented by Hassan A. Kingravi

Face Recognition Using Laplacianfaces He et al. (IEEE Trans PAMI, 2005) presented by Hassan A. Kingravi Face Recognition Using Laplacianfaces He et al. (IEEE Trans PAMI, 2005) presented by Hassan A. Kingravi Overview Introduction Linear Methods for Dimensionality Reduction Nonlinear Methods and Manifold

More information

Discriminative Direction for Kernel Classifiers

Discriminative Direction for Kernel Classifiers Discriminative Direction for Kernel Classifiers Polina Golland Artificial Intelligence Lab Massachusetts Institute of Technology Cambridge, MA 02139 polina@ai.mit.edu Abstract In many scientific and engineering

More information

Reading. 3. Image processing. Pixel movement. Image processing Y R I G Q

Reading. 3. Image processing. Pixel movement. Image processing Y R I G Q Reading Jain, Kasturi, Schunck, Machine Vision. McGraw-Hill, 1995. Sections 4.-4.4, 4.5(intro), 4.5.5, 4.5.6, 5.1-5.4. 3. Image processing 1 Image processing An image processing operation typically defines

More information

Lecture 8: Interest Point Detection. Saad J Bedros

Lecture 8: Interest Point Detection. Saad J Bedros #1 Lecture 8: Interest Point Detection Saad J Bedros sbedros@umn.edu Review of Edge Detectors #2 Today s Lecture Interest Points Detection What do we mean with Interest Point Detection in an Image Goal:

More information

Foliations of hyperbolic space by constant mean curvature surfaces sharing ideal boundary

Foliations of hyperbolic space by constant mean curvature surfaces sharing ideal boundary Foliations of hyperbolic space by constant mean curvature surfaces sharing ideal boundary David Chopp and John A. Velling December 1, 2003 Abstract Let γ be a Jordan curve in S 2, considered as the ideal

More information

Space-Variant Computer Vision: A Graph Theoretic Approach

Space-Variant Computer Vision: A Graph Theoretic Approach p.1/65 Space-Variant Computer Vision: A Graph Theoretic Approach Leo Grady Cognitive and Neural Systems Boston University p.2/65 Outline of talk Space-variant vision - Why and how of graph theory Anisotropic

More information

CSE 291. Assignment Spectral clustering versus k-means. Out: Wed May 23 Due: Wed Jun 13

CSE 291. Assignment Spectral clustering versus k-means. Out: Wed May 23 Due: Wed Jun 13 CSE 291. Assignment 3 Out: Wed May 23 Due: Wed Jun 13 3.1 Spectral clustering versus k-means Download the rings data set for this problem from the course web site. The data is stored in MATLAB format as

More information

Harmonic Flow for Histogram Matching

Harmonic Flow for Histogram Matching Harmonic Flow for Histogram Matching Thomas Batard and Marcelo Bertalmío Department of Information and Communication Technologies University Pompeu Fabra, Barcelona, Spain {thomas.batard,marcelo.bertalmio}@upf.edu

More information

ENERGY METHODS IN IMAGE PROCESSING WITH EDGE ENHANCEMENT

ENERGY METHODS IN IMAGE PROCESSING WITH EDGE ENHANCEMENT ENERGY METHODS IN IMAGE PROCESSING WITH EDGE ENHANCEMENT PRASHANT ATHAVALE Abstract. Digital images are can be realized as L 2 (R 2 objects. Noise is introduced in a digital image due to various reasons.

More information

Shape of Gaussians as Feature Descriptors

Shape of Gaussians as Feature Descriptors Shape of Gaussians as Feature Descriptors Liyu Gong, Tianjiang Wang and Fang Liu Intelligent and Distributed Computing Lab, School of Computer Science and Technology Huazhong University of Science and

More information

Heat Equations on Vector Bundles - Application to Color Image Regularization

Heat Equations on Vector Bundles - Application to Color Image Regularization Heat Equations on Vector Bundles - Application to Color Image Regularization Thomas Batard To cite this version: Thomas Batard. Heat Equations on Vector Bundles - Application to Color Image Regularization..

More information

Riemannian Drums, Anisotropic Curve Evolution and Segmentation

Riemannian Drums, Anisotropic Curve Evolution and Segmentation Riemannian Drums, Anisotropic Curve Evolution and Segmentation Jayant Shah * Mathematics Department, Northeastern University, Boston, MA 02115. * This work was partially supported by PHS Grant 2-R01 NS34189

More information

carroll/notes/ has a lot of good notes on GR and links to other pages. General Relativity Philosophy of general

carroll/notes/ has a lot of good notes on GR and links to other pages. General Relativity Philosophy of general http://pancake.uchicago.edu/ carroll/notes/ has a lot of good notes on GR and links to other pages. General Relativity Philosophy of general relativity. As with any major theory in physics, GR has been

More information

Lecture 6: Edge Detection. CAP 5415: Computer Vision Fall 2008

Lecture 6: Edge Detection. CAP 5415: Computer Vision Fall 2008 Lecture 6: Edge Detection CAP 5415: Computer Vision Fall 2008 Announcements PS 2 is available Please read it by Thursday During Thursday lecture, I will be going over it in some detail Monday - Computer

More information

Relativistic Mechanics

Relativistic Mechanics Physics 411 Lecture 9 Relativistic Mechanics Lecture 9 Physics 411 Classical Mechanics II September 17th, 2007 We have developed some tensor language to describe familiar physics we reviewed orbital motion

More information

IS&T's 2000 PICS Conference

IS&T's 2000 PICS Conference I&T's PIC Conference I&T's PIC Conference Copyright, I&T On Color Edge Detection Reiner Lenz Institute cience and Technology Campus Norrk oping, Link oping University, Bredgatan E-617 Norrk oping weden

More information

Generalized Newton-Type Method for Energy Formulations in Image Processing

Generalized Newton-Type Method for Energy Formulations in Image Processing Generalized Newton-Type Method for Energy Formulations in Image Processing Leah Bar and Guillermo Sapiro Department of Electrical and Computer Engineering University of Minnesota Outline Optimization in

More information

Spectral Processing. Misha Kazhdan

Spectral Processing. Misha Kazhdan Spectral Processing Misha Kazhdan [Taubin, 1995] A Signal Processing Approach to Fair Surface Design [Desbrun, et al., 1999] Implicit Fairing of Arbitrary Meshes [Vallet and Levy, 2008] Spectral Geometry

More information

Diffusion Geometries, Diffusion Wavelets and Harmonic Analysis of large data sets.

Diffusion Geometries, Diffusion Wavelets and Harmonic Analysis of large data sets. Diffusion Geometries, Diffusion Wavelets and Harmonic Analysis of large data sets. R.R. Coifman, S. Lafon, MM Mathematics Department Program of Applied Mathematics. Yale University Motivations The main

More information

Contribution from: Springer Verlag Berlin Heidelberg 2005 ISBN

Contribution from: Springer Verlag Berlin Heidelberg 2005 ISBN Contribution from: Mathematical Physics Studies Vol. 7 Perspectives in Analysis Essays in Honor of Lennart Carleson s 75th Birthday Michael Benedicks, Peter W. Jones, Stanislav Smirnov (Eds.) Springer

More information

Geometry-based Image Segmentation Using Anisotropic Diffusion TR November, 1992

Geometry-based Image Segmentation Using Anisotropic Diffusion TR November, 1992 Geometry-based Image Segmentation Using Anisotropic Diffusion TR92-039 November, 1992 Ross T. Whitaker Stephen M. Pizer Medical Image Display Group Department of Computer Science Department of Radiation

More information

NONLINEAR DIFFUSION PDES

NONLINEAR DIFFUSION PDES NONLINEAR DIFFUSION PDES Erkut Erdem Hacettepe University March 5 th, 0 CONTENTS Perona-Malik Type Nonlinear Diffusion Edge Enhancing Diffusion 5 References 7 PERONA-MALIK TYPE NONLINEAR DIFFUSION The

More information

Shape Matching: A Metric Geometry Approach. Facundo Mémoli. CS 468, Stanford University, Fall 2008.

Shape Matching: A Metric Geometry Approach. Facundo Mémoli. CS 468, Stanford University, Fall 2008. Shape Matching: A Metric Geometry Approach Facundo Mémoli. CS 468, Stanford University, Fall 2008. 1 GH: definition d GH (X, Y ) = inf Z,f,g dz H(f(X),g(Y )) 2 Remember the expression: For maps φ : X Y,

More information

has a lot of good notes on GR and links to other pages. General Relativity Philosophy of general relativity.

has a lot of good notes on GR and links to other pages. General Relativity Philosophy of general relativity. http://preposterousuniverse.com/grnotes/ has a lot of good notes on GR and links to other pages. General Relativity Philosophy of general relativity. As with any major theory in physics, GR has been framed

More information

Lecture 8: Interest Point Detection. Saad J Bedros

Lecture 8: Interest Point Detection. Saad J Bedros #1 Lecture 8: Interest Point Detection Saad J Bedros sbedros@umn.edu Last Lecture : Edge Detection Preprocessing of image is desired to eliminate or at least minimize noise effects There is always tradeoff

More information

Statistical Machine Learning from Data

Statistical Machine Learning from Data Samy Bengio Statistical Machine Learning from Data 1 Statistical Machine Learning from Data Support Vector Machines Samy Bengio IDIAP Research Institute, Martigny, Switzerland, and Ecole Polytechnique

More information

Principles of Riemannian Geometry in Neural Networks

Principles of Riemannian Geometry in Neural Networks Principles of Riemannian Geometry in Neural Networks Michael Hauser, Asok Ray Pennsylvania State University Presented by Chenyang Tao Nov 16, 2018 Brief Summary Goal This study deals with neural networks

More information

MTH4101 CALCULUS II REVISION NOTES. 1. COMPLEX NUMBERS (Thomas Appendix 7 + lecture notes) ax 2 + bx + c = 0. x = b ± b 2 4ac 2a. i = 1.

MTH4101 CALCULUS II REVISION NOTES. 1. COMPLEX NUMBERS (Thomas Appendix 7 + lecture notes) ax 2 + bx + c = 0. x = b ± b 2 4ac 2a. i = 1. MTH4101 CALCULUS II REVISION NOTES 1. COMPLEX NUMBERS (Thomas Appendix 7 + lecture notes) 1.1 Introduction Types of numbers (natural, integers, rationals, reals) The need to solve quadratic equations:

More information

Topics. CS Advanced Computer Graphics. Differential Geometry Basics. James F. O Brien. Vector and Tensor Fields.

Topics. CS Advanced Computer Graphics. Differential Geometry Basics. James F. O Brien. Vector and Tensor Fields. CS 94-3 Advanced Computer Graphics Differential Geometry Basics James F. O Brien Associate Professor U.C. Berkeley Topics Vector and Tensor Fields Divergence, curl, etc. Parametric Curves Tangents, curvature,

More information

B553 Lecture 3: Multivariate Calculus and Linear Algebra Review

B553 Lecture 3: Multivariate Calculus and Linear Algebra Review B553 Lecture 3: Multivariate Calculus and Linear Algebra Review Kris Hauser December 30, 2011 We now move from the univariate setting to the multivariate setting, where we will spend the rest of the class.

More information

Nonlinear diffusion filtering on extended neighborhood

Nonlinear diffusion filtering on extended neighborhood Applied Numerical Mathematics 5 005) 1 11 www.elsevier.com/locate/apnum Nonlinear diffusion filtering on extended neighborhood Danny Barash Genome Diversity Center, Institute of Evolution, University of

More information

(Quantum) Fields on Causal Sets

(Quantum) Fields on Causal Sets (Quantum) Fields on Causal Sets Michel Buck Imperial College London July 31, 2013 1 / 32 Outline 1. Causal Sets: discrete gravity 2. Continuum-Discrete correspondence: sprinklings 3. Relativistic fields

More information

Tensor Visualization. CSC 7443: Scientific Information Visualization

Tensor Visualization. CSC 7443: Scientific Information Visualization Tensor Visualization Tensor data A tensor is a multivariate quantity Scalar is a tensor of rank zero s = s(x,y,z) Vector is a tensor of rank one v = (v x,v y,v z ) For a symmetric tensor of rank 2, its

More information

Natural Evolution Strategies for Direct Search

Natural Evolution Strategies for Direct Search Tobias Glasmachers Natural Evolution Strategies for Direct Search 1 Natural Evolution Strategies for Direct Search PGMO-COPI 2014 Recent Advances on Continuous Randomized black-box optimization Thursday

More information

Lecture 4: Harmonic forms

Lecture 4: Harmonic forms Lecture 4: Harmonic forms Jonathan Evans 29th September 2010 Jonathan Evans () Lecture 4: Harmonic forms 29th September 2010 1 / 15 Jonathan Evans () Lecture 4: Harmonic forms 29th September 2010 2 / 15

More information

Connection of Local Linear Embedding, ISOMAP, and Kernel Principal Component Analysis

Connection of Local Linear Embedding, ISOMAP, and Kernel Principal Component Analysis Connection of Local Linear Embedding, ISOMAP, and Kernel Principal Component Analysis Alvina Goh Vision Reading Group 13 October 2005 Connection of Local Linear Embedding, ISOMAP, and Kernel Principal

More information

Lagrangian for Central Potentials

Lagrangian for Central Potentials Physics 411 Lecture 2 Lagrangian for Central Potentials Lecture 2 Physics 411 Classical Mechanics II August 29th 2007 Here we will review the Lagrange formulation in preparation for the study of the central

More information

Covariance Tracking Algorithm on Bilateral Filtering under Lie Group Structure Yinghong Xie 1,2,a Chengdong Wu 1,b

Covariance Tracking Algorithm on Bilateral Filtering under Lie Group Structure Yinghong Xie 1,2,a Chengdong Wu 1,b Applied Mechanics and Materials Online: 014-0-06 ISSN: 166-748, Vols. 519-50, pp 684-688 doi:10.408/www.scientific.net/amm.519-50.684 014 Trans Tech Publications, Switzerland Covariance Tracking Algorithm

More information

Scott Hughes 12 May Massachusetts Institute of Technology Department of Physics Spring 2005

Scott Hughes 12 May Massachusetts Institute of Technology Department of Physics Spring 2005 Scott Hughes 12 May 2005 24.1 Gravity? Massachusetts Institute of Technology Department of Physics 8.022 Spring 2005 Lecture 24: A (very) brief introduction to general relativity. The Coulomb interaction

More information

The Theorem of Gauß-Bonnet in Complex Analysis 1

The Theorem of Gauß-Bonnet in Complex Analysis 1 The Theorem of Gauß-Bonnet in Complex Analysis 1 Otto Forster Abstract. The theorem of Gauß-Bonnet is interpreted within the framework of Complex Analysis of one and several variables. Geodesic triangles

More information

Quasi-local Mass in General Relativity

Quasi-local Mass in General Relativity Quasi-local Mass in General Relativity Shing-Tung Yau Harvard University For the 60th birthday of Gary Horowtiz U. C. Santa Barbara, May. 1, 2015 This talk is based on joint work with Po-Ning Chen and

More information

Curved spacetime and general covariance

Curved spacetime and general covariance Chapter 7 Curved spacetime and general covariance In this chapter we generalize the discussion of preceding chapters to extend covariance to more general curved spacetimes. 219 220 CHAPTER 7. CURVED SPACETIME

More information

Abstract Surfaces in Image Processing

Abstract Surfaces in Image Processing Abstract Surfaces in Image Processing E. Meinhardt-Llopis, G. Facciolo March 27, 2014 Abstract In this article we review the typical uses of Riemannian geometry in image processing and we propose new applications

More information

CS 468, Lecture 11: Covariant Differentiation

CS 468, Lecture 11: Covariant Differentiation CS 468, Lecture 11: Covariant Differentiation Adrian Butscher (scribe: Ben Mildenhall) May 6, 2013 1 Introduction We have talked about various extrinsic and intrinsic properties of surfaces. Extrinsic

More information

Scientific Data Computing: Lecture 3

Scientific Data Computing: Lecture 3 Scientific Data Computing: Lecture 3 Benson Muite benson.muite@ut.ee 23 April 2018 Outline Monday 10-12, Liivi 2-207 Monday 12-14, Liivi 2-205 Topics Introduction, statistical methods and their applications

More information

Information geometry for bivariate distribution control

Information geometry for bivariate distribution control Information geometry for bivariate distribution control C.T.J.Dodson + Hong Wang Mathematics + Control Systems Centre, University of Manchester Institute of Science and Technology Optimal control of stochastic

More information

A BAYESIAN APPROACH TO NONLINEAR DIFFUSION BASED ON A LAPLACIAN PRIOR FOR IDEAL IMAGE GRADIENT

A BAYESIAN APPROACH TO NONLINEAR DIFFUSION BASED ON A LAPLACIAN PRIOR FOR IDEAL IMAGE GRADIENT A BAYESIAN APPROACH TO NONLINEAR DIFFUSION BASED ON A LAPLACIAN PRIOR FOR IDEAL IMAGE GRADIENT Aleksandra Pižurica, Iris Vanhamel, Hichem Sahli, Wilfried Philips and Antonis Katartzis Ghent University,

More information

Partial Differential Equations and Image Processing. Eshed Ohn-Bar

Partial Differential Equations and Image Processing. Eshed Ohn-Bar Partial Differential Equations and Image Processing Eshed Ohn-Bar OBJECTIVES In this presentation you will 1) Learn what partial differential equations are and where do they arise 2) Learn how to discretize

More information

The Uses of Differential Geometry in Finance

The Uses of Differential Geometry in Finance The Uses of Differential Geometry in Finance Andrew Lesniewski Bloomberg, November 21 2005 The Uses of Differential Geometry in Finance p. 1 Overview Joint with P. Hagan and D. Woodward Motivation: Varadhan

More information

A Pseudo-distance for Shape Priors in Level Set Segmentation

A Pseudo-distance for Shape Priors in Level Set Segmentation O. Faugeras, N. Paragios (Eds.), 2nd IEEE Workshop on Variational, Geometric and Level Set Methods in Computer Vision, Nice, 2003. A Pseudo-distance for Shape Priors in Level Set Segmentation Daniel Cremers

More information

Dimension Reduction Techniques. Presented by Jie (Jerry) Yu

Dimension Reduction Techniques. Presented by Jie (Jerry) Yu Dimension Reduction Techniques Presented by Jie (Jerry) Yu Outline Problem Modeling Review of PCA and MDS Isomap Local Linear Embedding (LLE) Charting Background Advances in data collection and storage

More information

M3/4A16. GEOMETRICAL MECHANICS, Part 1

M3/4A16. GEOMETRICAL MECHANICS, Part 1 M3/4A16 GEOMETRICAL MECHANICS, Part 1 (2009) Page 1 of 5 UNIVERSITY OF LONDON Course: M3/4A16 Setter: Holm Checker: Gibbons Editor: Chen External: Date: January 27, 2008 BSc and MSci EXAMINATIONS (MATHEMATICS)

More information

Emergent space-time and gravity in the IIB matrix model

Emergent space-time and gravity in the IIB matrix model Emergent space-time and gravity in the IIB matrix model Harold Steinacker Department of physics Veli Losinj, may 2013 Geometry and physics without space-time continuum aim: (toy-?) model for quantum theory

More information

THEODORE VORONOV DIFFERENTIAL GEOMETRY. Spring 2009

THEODORE VORONOV DIFFERENTIAL GEOMETRY. Spring 2009 [under construction] 8 Parallel transport 8.1 Equation of parallel transport Consider a vector bundle E B. We would like to compare vectors belonging to fibers over different points. Recall that this was

More information

Color Image Enhancement via Chromaticity Diffusion

Color Image Enhancement via Chromaticity Diffusion IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 10, NO. 5, MAY 2001 701 Color Image Enhancement via Chromaticity Diffusion Bei Tang, Guillermo Sapiro, Member, IEEE, and Vicent Caselles Abstract A novel approach

More information

Introduction and Vectors Lecture 1

Introduction and Vectors Lecture 1 1 Introduction Introduction and Vectors Lecture 1 This is a course on classical Electromagnetism. It is the foundation for more advanced courses in modern physics. All physics of the modern era, from quantum

More information

Statistical and Computational Analysis of Locality Preserving Projection

Statistical and Computational Analysis of Locality Preserving Projection Statistical and Computational Analysis of Locality Preserving Projection Xiaofei He xiaofei@cs.uchicago.edu Department of Computer Science, University of Chicago, 00 East 58th Street, Chicago, IL 60637

More information

dy = f( x) dx = F ( x)+c = f ( x) dy = f( x) dx

dy = f( x) dx = F ( x)+c = f ( x) dy = f( x) dx Antiderivatives and The Integral Antiderivatives Objective: Use indefinite integral notation for antiderivatives. Use basic integration rules to find antiderivatives. Another important question in calculus

More information

2.1 The metric and and coordinate transformations

2.1 The metric and and coordinate transformations 2 Cosmology and GR The first step toward a cosmological theory, following what we called the cosmological principle is to implement the assumptions of isotropy and homogeneity withing the context of general

More information

Consensus Algorithms for Camera Sensor Networks. Roberto Tron Vision, Dynamics and Learning Lab Johns Hopkins University

Consensus Algorithms for Camera Sensor Networks. Roberto Tron Vision, Dynamics and Learning Lab Johns Hopkins University Consensus Algorithms for Camera Sensor Networks Roberto Tron Vision, Dynamics and Learning Lab Johns Hopkins University Camera Sensor Networks Motes Small, battery powered Embedded camera Wireless interface

More information

Physics 6303 Lecture 2 August 22, 2018

Physics 6303 Lecture 2 August 22, 2018 Physics 6303 Lecture 2 August 22, 2018 LAST TIME: Coordinate system construction, covariant and contravariant vector components, basics vector review, gradient, divergence, curl, and Laplacian operators

More information

Statistical Machine Learning from Data

Statistical Machine Learning from Data January 17, 2006 Samy Bengio Statistical Machine Learning from Data 1 Statistical Machine Learning from Data Other Artificial Neural Networks Samy Bengio IDIAP Research Institute, Martigny, Switzerland,

More information

Photometric Stereo: Three recent contributions. Dipartimento di Matematica, La Sapienza

Photometric Stereo: Three recent contributions. Dipartimento di Matematica, La Sapienza Photometric Stereo: Three recent contributions Dipartimento di Matematica, La Sapienza Jean-Denis DUROU IRIT, Toulouse Jean-Denis DUROU (IRIT, Toulouse) 17 December 2013 1 / 32 Outline 1 Shape-from-X techniques

More information

Edge Detection in Computer Vision Systems

Edge Detection in Computer Vision Systems 1 CS332 Visual Processing in Computer and Biological Vision Systems Edge Detection in Computer Vision Systems This handout summarizes much of the material on the detection and description of intensity

More information

Tensors, and differential forms - Lecture 2

Tensors, and differential forms - Lecture 2 Tensors, and differential forms - Lecture 2 1 Introduction The concept of a tensor is derived from considering the properties of a function under a transformation of the coordinate system. A description

More information

Generalized Gradient on Vector Bundle - Application to Image Denoising

Generalized Gradient on Vector Bundle - Application to Image Denoising Generalized Gradient on Vector Bundle - Application to Image Denoising Thomas Batard, arcelo Bertalmío To cite this version: Thomas Batard, arcelo Bertalmío. Generalized Gradient on Vector Bundle - Application

More information

Digital Image Processing ERRATA. Wilhelm Burger Mark J. Burge. An algorithmic introduction using Java. Second Edition. Springer

Digital Image Processing ERRATA. Wilhelm Burger Mark J. Burge. An algorithmic introduction using Java. Second Edition. Springer Wilhelm Burger Mark J. Burge Digital Image Processing An algorithmic introduction using Java Second Edition ERRATA Springer Berlin Heidelberg NewYork Hong Kong London Milano Paris Tokyo 5 Filters K K No

More information

Lecture No 1 Introduction to Diffusion equations The heat equat

Lecture No 1 Introduction to Diffusion equations The heat equat Lecture No 1 Introduction to Diffusion equations The heat equation Columbia University IAS summer program June, 2009 Outline of the lectures We will discuss some basic models of diffusion equations and

More information

Chapter 7 Curved Spacetime and General Covariance

Chapter 7 Curved Spacetime and General Covariance Chapter 7 Curved Spacetime and General Covariance In this chapter we generalize the discussion of preceding chapters to extend covariance to more general curved spacetimes. 145 146 CHAPTER 7. CURVED SPACETIME

More information

Random walks and anisotropic interpolation on graphs. Filip Malmberg

Random walks and anisotropic interpolation on graphs. Filip Malmberg Random walks and anisotropic interpolation on graphs. Filip Malmberg Interpolation of missing data Assume that we have a graph where we have defined some (real) values for a subset of the nodes, and that

More information

Faculty of Engineering, Mathematics and Science School of Mathematics

Faculty of Engineering, Mathematics and Science School of Mathematics Faculty of Engineering, Mathematics and Science School of Mathematics GROUPS Trinity Term 06 MA3: Advanced Calculus SAMPLE EXAM, Solutions DAY PLACE TIME Prof. Larry Rolen Instructions to Candidates: Attempt

More information