Image segmentation: A competitive approach
Abstract
Image segmentation is viewed as an optimization problem, where the goal is to construct a partition such that nearby pixels belonging to the same region are as similar as possible, and nearby pixels belonging to different regions are as different as possible. Such constraints, which involve relations among nearby pixels, can be handled by formulating local hypotheses, e.g. that a small contiguous set of pixels should be grouped together, or that a small set of contiguous edges should be linked together. These hypotheses first compete at the local level; the winners at this level go on to compete at the global level. Competition at the global level is carried out by sorting the hypotheses based on a local cost function, and then applying them one at a time, in sort order, as constraints on the final segmentation.
Degree
Ph.D.
Advisors
Tenorio, Purdue University.
Subject Area
Electrical engineering|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.