为精确、全面地描述中医望诊面色,提出一种利用光谱反射率进行面部各像素点颜色表述的方法.首先基于有限维模型对面部光谱反射率基函数进行分析;然后设计了面部多光谱照明信息采集系统,从光谱角度确定适用于面部光谱反射率复原的光照光谱组合,并基于平场校正原理对图像进行均匀性校正,解决了图像采集过程中光照不均匀问题;最后建立多光谱图像数据与对应光谱反射率之间的函数关系模型.实验结果表明,利用该方法进行面部光谱复原,仅需较少参数即可达到更好的复原效果.
A spectral reflectance recovery method based on multiplex illumination was proposed to be aimed at describing the facial complexion of each pixel more accurately. The main procedures are as follows: basis functions of facial spectral reflectance were analyzed based on the finite dimension model; the multispectral illumination system for facial image acquisition was designed, and the spectrum composition of the illumination applicable for the spectral reflectance recovery was determined; the influence of uneven illumination was reduced by correcting the image data based on flat-fielding; the function model between multi-spectral image data and the basis function coefficients was constructed to achieve spectral reflectance recovery. The results show that this method can achieve relatively high accuracy of facial spectral reflectance recovery with fewer parameters.