In the talk, we will discuss three of the key mathematical issues in computer vision. First using the theory of invariant curvature driven flows, we will describe algorithms for image denoising and smoothing. Second, using optimal mass transport, we will derive new ways for dealing with the problems of elastic registration, mesh generation, and dynamic textures. Finally, we will give a new control formulation of the interactive segmentation in which user defined movements are made an explicit part of the cost functional.

Numerous examples will be included from the medical imaging side, and the talk itself will be directed at a broad audience interested in modern methods in image processing and computer vision.