Date: Thursday April 17, 2008
Place: CSE E305
Time: 4:00 pm to 5:00 pm
Speaker: Hang Yu
Topic:
      String Kernels for Matching Seriated Graphs
Source: Link
Abstract:
Graph seriation allows the nodes of a graph to be placed in a string order, and then matched using string alignment algorithms. Prior work has used Bayesian methods to derive the string edit costs required in matching. The aim in this paper is to demonstrate how the matching of seriated graphs can be kernelised. To do this we make use of string kernels. We illustrate that the graph edit distances computed using the string kernel can be used for graph clustering.

Date: Thursday April 4, 2008
Place: CSE E305
Time: 4:00 pm to 5:00 pm
Speaker: Ozlem Subakan
Topic:
      Feature preserving smoothing and segmentation via convolution with a spatially varying kernel
Source:
Abstract:
Many computer vision and image processing tasks require the preservation of local discontinuities, terminations and bifurcations. Denoising with feature preservation is a challenging task and in this paper, we present a novel technique for preserving complex oriented structures such as junctions and corners present in images. This is achieved in a two stage process namely,(1) All image data are pre-processed to extract local orientation information using a steerable Gabor filter bank. The orientation distribution at each lattice point is then represented by a continuous mixture of Gaussians. The continuous mixture representation can be cast as the Laplace transform of the mixing density over the space of positive definite (covariance) matrices. This mixing density is assumed to be a parameterized distribution, namely, a mixture of Wisharts whose Laplace transform is evaluated in a closed form expression called the Rigaut type function, a scalar-valued function of the parameters of the Wishart distribution. Computation of the weights in the mixture of Wisharts is formulated as a sparse deconvolution problem. (2) The feature preserving denoising is then achieved via iterative convolution of the given image data with the Rigaut type function. We present experimental results on noisy data, real 2D images and 3D MRI data acquired from plant roots depicting bifurcating roots. Superior performance of our technique is depicted via comparison to the state-of-the-art anisotropic diffusion filter.

Date: Thursday February 28, 2008
Place: CSE E305
Time: 4:00 pm to 5:00 pm
Speaker: Prof. Anand Rangarajan
Topic: Hamilton-Jacobi Theory and the Calculus of Variations
Source: Link
Abstract:
We will disucss the relationships between Hamilton-Jacobi theory and the calculus of variations.

Date: Thursday February 14, 2008
Place: CSE E305
Time: 4:00 pm to 5:00 pm
Speaker: Prof. Jeffrey Ho
Topic: Using Galois theory to prove structure from motion algorithms are optimal
Source: Link
Abstract:
In this talk, we will discuss a CVPR2007 paper by Nister, Hartley and Stewenius on proving the optimality of the algorithms for solving two structure from motion problems. Specifically, the two problems are 1) five point calibrated relative orientation and 2) L2-optimal two-view triangulation. These two problems are known to require the solutions of a tenth degree and a sixth degree polynomials, respectively. This paper shows that, using Galois theory, it is not possible to invent a general algorithm for each problem by solving a polynomial of lesser degree.

Date: Thursday February 7, 2008
Place: CSE E305
Time: 4:00 pm to 5:00 pm
Speaker: Ritwik Kumar
Topic: Illumination in Face Recognition
Source: none
Abstract:
I will be presenting a survey of techniques that deal with illumination in context of face recognition and then present some of our results.

Date: Thursday January 31, 2008
Place: CSE E305
Time: 4:00 pm to 5:00 pm
Speaker: Venkatakrishnan Ramaswamy
Topic: On feedforward networks of neurons
Source: none
Abstract:
A feedforward network of neurons is one where connections between the neurons do not form a directed cycle. We motivate the problem, discuss some recent results and open problems.

Date: Thursday January 24, 2008
Place: CSE E305
Time: 4:00 pm to 5:00 pm
Speaker: Ajit Rajwade
Topic: Density Estimation
Source:
     Nonlinear Mean Shift for Clustering over Analytic Manifolds
     Discontinuity Preserving Filtering over Analytic Manifolds
Abstract:
Ajit will discuss both papers.