A color space study for skin lesion segmentation

Abstract

The objective of this research is to identify the most suitable color model for segmentation of Eczema skin lesions. Eczema is a type of Atopic Dermatitis that is diagnosed by the dermatologists by visual inspection hence by subjective assessment. For segmentation, K-means clustering has been used. Four color spaces, i.e. HSI, CMY, YCbCr and CIELAB are used for comparisons. The final conclusion is made on the basis of the results for sensitivity, specificity and accuracy for each technique. It has been observed that the color channel “H” of HSI; followed by “a” of CIELab; gives better segmentation results.