Semi-automatic Detection of Skin Malformations by Analysis of Spectral Images

摘要

The multi-spectral imaging technique to reveal skin malformations has been described in this work. Four spectral images taken at polarized monochromatic LED illumination (450nm, 545nm, 660nm and 940 nm) and polarized white LED light imaged by CMOS sensor via cross-oriented polarizing filter were analyzed to calculate chromophore maps. The algorithm based on skin color analysis and user-defined threshold selection allows highlighting of skin areas with predefined chromophore concentration semi-automatically. Preliminary results of clinical tests are presented.