An active contour model using nonlinear prior shape

Abstract

A novel level set image segmentation method using the prior shape is proposed in this paper in view of the problem which occurs when the existing level set method using the prior shape segmented the images with strong noise, weak boundary or complicated background. The kernel principal component analysis is used in this method to decrease the dimensions of the training samples and extract the principal component as the prior shape to guide the segmentation. Then the novel method does expansion on the mean shape which is used as the initial contour to effectively solve the determined problem of the initial contour of the curve evolution. The variational level set method is adopted in the novel method, and the local binary fitting model and the priori shape energy term is combined. Experiments show that the novel method has better segmentation results and higher segmentation efficiency on the images with strong noise, weak boundary or complicated background.