Robust fuzzy snakes

Flexible closed deformable contours extracted from images to represent object outlines are often called snakes. The paper explores and extends the use of snakes to extract contour features from images for modeling and pattern recognition. Many of the contour extraction algorithms have a common formulation as a constrained optimization problem. One version, called the constrained fuzzy clustering algorithm, links fuzzy clustering and snakes-the resulting contour being called a fuzzy snake. These extraction procedures are vulnerable to noise, which makes their application to noisy sensor images problematic. By applying known robust fuzzy clustering algorithms, more noise resistant snakes can be derived thereby extending contour feature extraction to sensor images.