Skin Image Synthesis Based on Pigment Concentration Distribution

Abstract

An image synthesis algorithm based on three dominant pigments is proposed in this paper to reproduce high fidelity skin image. Combined the Lambert Beer law with skin multiple layers structure, the mathematical relationship between skin pigments concentration and image pixel value is firstly introduced. To eliminate the influence of surface reflection component, the difference of each color channel in log space is applied to form 2D mixture signals. The ICA algorithm is then followed to obtain melanin and hemoglobin concentration distribution. Successively, carotene concentration distribution is given by linear regression. The three pigment concentration coefficients will dominate skin image synthesis. The experiments demonstrate the reproduced skin images based on change of pigment concentration coefficients. In addition, the comparisons with manual data show the effectiveness of proposed algorithm.