Event Location: NSH 1507
Bio: Larry Zitnick received his PhD degree in robotics from Carnegie Mellon University in 2003. His thesis focused on algorithms for efficiently computing conditional probabilities in large-problem domains. Previously, his work centered on stereo vision, including cooperative and parallel algorithms, as well as the commercial development of a portable 3D camera. Currently, he is a researcher at the Interactive Visual Media group at Microsoft Research. His latest work includes video interpolation, object recognition, and computational photography.

Abstract: In this talk I discuss three projects related to low-level image manipulations. First, I describe an algorithm for automatic estimation and removal of noise using piecewise smooth image models. The generative model assumes an image is formed from a set of affine smooth regions. Next, I discuss a method for demosaicing Raw images into RGB images using a two color prior. This prior makes the assumption that only two colors exist within a small neighborhood around a pixel. My final topic explores gradient domain problems, such as sharpening, colorization and blending, in which data constraints are placed on an image, as well as the standard gradient constraints. I show that this set of problems can also be solved using fast Poisson solvers.