Abstract

Introduction to Image Segmentation: Part 1b

Yuri Boykov
University of Western Ontario

(Parts 1 and 2) This lecture discusses different formulations for the binary segmentation problem. We compare a number of standard models, their motivation, and the corresponding optimization algorithms. While focusing on discrete approaches, we compare them with standard continuous methods. Keywords: dynamic programming, shortest–paths, objective functions (regularization energies), region vs boundary terms, graph cuts vs level sets vs active contours and other methods, geometry vs physics, integral geometry vs differential geometry, discrete vs continuous models, interactive and unsupervised segmentation. Set functions. Submodularity