Image-based skin color and texture analysis/synthesis by extracting hemoglobin and melanin information in the skin
Norimichi Tsumura* Nobutoshi Ojima† Kayoko Sato* Mitsuhiro Shiraishi† Hideto Shimizu*
Hirohide Nabeshima† Syuuichi Akazaki† Kimihiko Hori† Yoichi Miyake*
*Chiba Univerisity †Kao Corporation
To appear in ACM Transactions on Graphics Vol.22 No.3,
pp.770-779, SIGGRAPH 2003
This paper proposes an E-cosmetic function for digital images based on physics and physiologically-based image processing. A practical skin color and texture analysis/synthesis technique is introduced for this E-cosmetic function. Shading on the face is removed by a simple color vector analysis in the optical density domain as an inverse lighting technique. The image without shading is analyzed by a previously introduced technique that extracts hemoglobin and melanin components by independent component analysis. Experimental results using UV-B irradiation and the application of methyl nicotinate on the arms support the physiological validity of the analysis and the effectiveness of the proposed shading removal. We synthesized the way facial images changed due to tanning or alcohol consumption, and compared the synthesized images with images of actual changes in skin color. The comparison shows an excellent match between the synthesized and actual images of changes due to tanning and alcohol consumption. We also proposed a technique to synthesize the change of texture in pigment due to aging or the application of cosmetics. The pyramid-based texture analysis/synthesis technique was used for the spatial processing of texture. Using the proposed technique, we could realistically change the skin color and texture of a 50 year-old woman to that of a 20 year-old woman.