光强倒数色度空间的彩色人脸光照预处理

为了解决彩色人脸识别中色彩信息易受光照影响的问题，提出一种基于光强倒数色度空间（IICS）的彩色人脸图像预处理方法。本方法首先将图像均匀地分割成子块；将每个图像块变换为 IICS 空间中的一个二维数据集，并根据数据集的线性分布特性估计图像块的光照颜色；然后对全部图像块的光照估计进行颜色直方图统计，根据直方图对分块估计的结果进行合并；最后，利用估计得到的光照和对角模型将图像光照校正到标准白光下，用于人脸识别。在 AR 和 FERET 人脸库上的实验表明，通过引入本光照预处理，有效增强了彩色人脸识别方法对光照变化的鲁棒性，提高了识别精度。

Illumination preprocessing of color face based on inverse-intensity chromaticity space

An image preprocessing method based on the inverse-intensity chromaticity space(IICS)theory is proposed to solve the problem that the color information is sensitive to illumination condition in a color face recognition method. This method evenly divides the image into blocks. Then,the pixels in each block are converted to a 2 dimensional data set in the IICS,which takes a linear distribution,and the illumination color of this block is estimated from this data set. After the illumination colors of all the blocks are derived,a color histogram is computed based on the estimated colors,and the colors are merged according to the histogram. Finally,illumination condition of the image is rectified to the canonical illumination using the estimated colors and the diagonal model,and this rectified image is used for recognition. Experiments on the AR and FERET face databases show that the robustness to illumination variation is effectively improved,and the recognition accuracy is increased by introducing the illumination preprocessing scheme into a color face recognition method.