SKIN HEMOGLOBIN AND MELANIN QUANTIFICATION ON MULTI-SPECTRAL IMAGES

摘要

In this paper, we propose and compare two differentrnapproaches for quantification of skin hemoglobin andrnmelanin on multi-spectral images. The first method isrnbased on non-negative matrix factorization (NMF) withrnmultiplicative update algorithm. The second methodrnis a Beer-Lambert law based model-fitting technique.rnQuantitative evaluation through graph-cut segmentationrnon melanoma indicates that model-fitting method obtainsrnmore accurate quantification than NMF.