Segmentation of Dermatoscopic Images Using Wavelet Transform
摘要
Wavelet-based segmentation provides a very effective technique for medical images. Automatic segmentation of skin lesions is the first step toward development of computer-aided diagnosis of melanoma. This paper provides an improved automated skin lesion segmentation method for color dermatoscopic images. A novel wavelet transform-based technique called “Segmentation of Dermatoscopic images using wavelet transform” (SDIWT) is proposed. One of the important advantages of wavelet transform is that it provides a precise and unifying framework for the analysis and characterization of a signal at different levels. SDIWT uses a perceptually uniform color space for segmentation. To reduce computational complexity for clustering, prominent pixels are selected. One-level decomposition of Daubechies wavelet function is used. LL1 sub-band of decomposition is utilized for clustering. Fuzzy c-means (FCM) clustering technique is used to find out clusters and their labels. Fuzzy entropy is used to decide number of clusters. The image pixels are classified to respective clusters based on minimum Euclidean distance. A post-processing noise-filtering stage is applied to improve the segmentation output. Edges are determined to specify the lesion from dermatoscopic images. One of the advantages of the proposed method is that it does not require to specify a priori information to segment a color region. The competence of the proposed method has been demonstrated by various experiments.