| 摘要 |
A digital photograph comprising identified skin parts and analyzing cutaneous lesions is analysed by first enhancing lesions in said identified skin parts then detecting hair patches then approximating localization of all lesions and finally identifying lesions pixels. The enhancement process may involve detecting skin complexion using common/averaged density estimation on a dominant channel extracted from skin pixels, boosting lesion pixels and suppressing skin pixels and combining dominant channel with the boosted pixels and the hair detection may involve semi-supervised k-clustering or spectral clustering. |
