Denoised Feature Extraction Based On The Tensor Voting Theory

Crest lines have a nice mathematical background. Crest lines, as curves on a surface along which the surface bends sharply, are powerful shape descriptors. Because crest lines and their subsets have numerous applications in image analysis, face recognition, smoothing, surface segmentation and hole-filling, robust extraction of crest lines is a very important issue.Recently, some algorithms for the detection of crest lines on dense triangle meshes have been presented. Reliable computations of discrete principal curvature measures on meshes are key to the detection of crest lines. In general, in order to estimate the principal curvature of a point, we need a local coordinate system, which are followed by all subsequent calculations. Owing to over-locality, it is inadequate to detect significant crest lines from a noise model. In this paper, a method is discussed for robust detection of crest lines. The method is based on the tensor voting theory. It classifies a feature into a corner, a sharp edge and a face. We have made improvements by incorporating the method with contextual information, the attributes of neighboring points. So it provides a basis for robustly detecting salient crest lines corresponding to potentially important features. Consequently, the algorithm is immune to noisy mesh. Comparative results indicate that our algorithm yields favorable detection results and is effective.