Boundary irregularity of skin lesions is of clinical significance for early detection of malignant melanomas, while structural components of lesion contour have clinical significance in particular. To extract effective features of structural components of skin lesions, an integrated approach is proposed in this paper using wavelet decompositions and inter-class distance analysis (Hausdorff Distance, HD for short) among wavelet subbands. Firstly, lesion contours are modeled as signatures with scale normalization to reach position and frequency resolution invariant. Energy distributions among different wavelet sub-bands are then analyzed to extract those significant levels with large discrimination capability. Based on significant subbands extracted, structural components of original contours are generated and irregularity descriptors at the multi-significant subbands are presented. Finally experimental results give effectiveness of proposed methods.