Wavelet-Based Approach for Skeleton Extraction

We propose a novel wavelet-based approach to extract skeleton of ribbon-like shapes based on a new concept called wavelet-based local modulus maxima symmetry (WLMMS). The development of the new approach benefits from the desirable properties of the constructed new wavelet function. Based on the proposed WLMMS, initial skeleton of the regular region of ribbon-like shape are computed. Special attention is given to development of a new amendment technique which is called interpolation compensation. It is used to to remove artifacts of the initial skeletons and compute the skeletons in the singular region of the shape. Experimental results show that the proposed approach is not only capable of extracting precisely the skeleton of the ribbonlike shape with the low computational cost, but also robust against noise. And the proposed algorithm is applicable to both binary and gray scale image, as most traditional methods fail to do.