Search on extracting silhouette features of objects in virtual reality

In this paper, a new method for detecting silhouette(outline) is proposed with the help of extremum of first-order directional derivatives and zero of second- order directional derivatives, which is based upon a general gradient model. Silhouette lines are extracted automatically from images by acquiring points of extremum of OF first-order directional derivatives vertical with silhouette lines. The procedure is accomplished in frequency domain with Fourier transform. In detecting silhouette, non-silhouette points are rejected by using values of first-order directional derivatives vertical with silhouette line. It is such an important characteristic that the method proposed in the paper does not need to choose the threshold. The experimental results show the. method is useful and practical, and precision detecting silhouette is good.