Abstract

Segmentation of imagery data has received much attention because of its potential application in several disciplines, e.g., medicine, remote-sensing, robot vision etc.. The image segmentation problem can be specified in the following way: let A be the index set for the pixel values for the whole image; we assume that there exists a partition on A, A = A₁ ⋃ A₂ ⋃…⋃ A_q such that A_i ≠ ϕ and A_i ∩ A_j = ϕ ∀ i ≠ j, where each A_i is homogeneous in some sense. Thus the problem is to find A₁, A₂,...,A_q. The segmentation problem has been approached using both structural and statistical procedures.

© 1985 Optical Society of America