A Novel Two-Step-Deformation Active Contour with Application to Facial Features Extraction
Zhang Hua, Yuanquan Wang · 2009
In allusion to the boundary leaking problem of traditional active contours when used to extract facial features, two kinds of local information are introduced to enhance the weak boundary, then to build new external force for active contours. The prior knowledge of the local shape of the facial features is incorporated into the evolution of active contour. The snake contour in each evolution step just serves as prediction, and then is ruled by the prior local shape; this way, the snake contour deforms via prediction-correction two steps. This two step-deformation (TSD) snake can utilize the a priori shape of objects effectively. The proposed strategies are evaluated on several images and the experimental results validate the high performance of this TSD snake for facial features segmentation.