利用物理模型合成皮肤图像

为了获得各种丰富和逼真的皮肤图像,在对皮肤结构和皮肤组成进行分析的基础上,提出了一种基于物理模型的鲁棒算法对输入的单张皮肤图像自动进行色素分离,并在此基础上实现了全局性和局部性的合成。输入的图像首先被分成N个子区域,并依次在每个子区域利用ICA算法提取分离向量,之后对得到的分离向量进行合理性验证。如果分离向量不合理,则丢弃,在下一个子区域重复上述过程。最后,从所有的合理分离向量中得到最终的分离向量,分离得到黑色素和血色素的密度分布图。通过改变色素的密度可以进行全局性合成,也可以利用密度信息选取感兴趣区域进行局部性合成。实验结果表明,该算法非常有效,可以对人体皮肤图像进行"电子化妆",重新生成更多逼真的合成图像。

Skin Image Synthesis Based on Physical Model

To produce various realistic skin images,a physical-model-based and robust algorithm is proposed in this paper to automatically separate single skin image into melanin and hemoglobin components based on the analysis of skin structure and compositions.Skin image can then be locally or globally synthesized based on pigment components separation.Firstly,we divide the input image into several sub-regions and ICA algorithm is successively performed in each sub-region to extract separation vectors.After testing the validity of separation vector,invalid separation vector is discarded and the valid are considered as candidate vectors.Repeat above operations and we obtain final separation vector from the collection of candidate vectors for obtaining melanin and hemoglobin component map.Based on separated component map,we can globally synthesize new skin image,or select ROI(Region of Interest)to locally synthesize.Our experiments show that the proposed algorithm is very effective and can act as an e-cosmetic function to reproduce realistic results.