Fast and accurate active contours for object boundary segmentation
Cheng‐Hung Chuang, Wen‐Nung Lie · 2002
We propose a new scheme of active contours for object boundary segmentation. The scheme is based on a greedy, search algorithm over a gradient vector flow (GVF) field. This method is less sensitive to initial contour conditions and accurate in approaching concave parts of object boundary. The basic ideas are to propagate edge information of object boundaries into a move spread areas near the initial contours by utilizing the GVF field and to speed up processing by applying a greedy search algorithm. A preprocessing procedure is also proposed to enhance the true gradient information in noisy images. Some artificial and real images are used for simulations and analysis. The results show accurate precision and fast speed for object boundary extraction.