An improved image segmentation method for melasma severity assessment

Melasma is a widely spread skin pigmentation disease and accurate assessments of the disease severity is crucial during its treatment. Recently, several computerized methods have been developed to overcome the shortcomings of the conventional clinical assessment method. As a key step in algorithm, image segmentation has extensive impacts on the accuracy of the assessment. Currently, the optimal hybrid thresholding (oHybrid) segmentation method that adaptively combines both local and global thresholding methods has obtained reasonable results in solving the melasma assessment problem. Nevertheless, the distance measure adopted in the oHybrid method is empirically selected and its influence is not well discussed. In this paper, a generalized distance measure is introduced and applied to improve the hybrid thresholding image segmentation method. The proposed method is tested on a data set of melasma patients to determine the lesion severity and it shows the best overall performance among the methods compared.