Graph-Cuts Technique For Melanoma Segmentation Over Different Color Spaces

Application of automated image analysis techniques for the detection, diagnosis and management melanoma continues to be an active research area globally. Although a lot of progress has been made on the study of different automated methods of melanoma segmentation, there is still need for further improvement. This paper presents a study on the use of graph-cuts technique for the segmentation of melanoma in clinical images over four different color spaces. The four color spaces considered in this study are RGB, HSV, HSI and HSL. Experimental results show that the use of graph-cuts technique over all the four color spaces are very promising as the average accuracy rate of 96.98%, average sensitivity rate of 89.68%, average specificity rate of 98.96%, average precision rate of 96.34% and average f-score rate of 93,51% are achieved.