| | |
Summary: Feature Weighted Active Contours for Image Segmentation
Bing Li and Scott T. Acton
Charles L. Brown Department of Electrical and Computer Engineering
University of Virginia, Charlottesville, VA 22904 USA
E-mail: {bingli, acton}@virginia.edu
Abstract
A novel external energy construction for active
contours is proposed in this paper. In contrast to the
standard approach of using linear filtering to smooth
the external energies and to avoid noise, we define
feature weighted active contours for extracting
features of interest without distortion. The advantages
of this innovation are demonstrated by examples and
comparisons with Gaussian filtered external energy.
Compared to the linear filtering approach, the feature
weighted active contours yield lower root mean
squared errors of contour position and improve upon
the ability to capture fine details in noisy images.
1. Introduction
Snakes [1], or active contours, have been widely
|
