基于方向扫描的图像光照补偿算法

Line scanning for illumination compensation of image

提出了一种基于二维图像信息的人脸光照补偿算法。基于人脸形状与球面相近的假设,首先沿人脸对称方向估算均匀光照下人脸图像灰度的统计信息,并对统计信息进行数据拟合,构建标准光照模型;然后估算非均匀光照下人脸图像的大致光照方向,并沿垂直光照方向对图像灰度进行统计分析;最后利用标准光照模型,结合线性和非线性变换,把非均匀光照下人脸图像调整到标准状态。在Yale B人脸库上的处理结果表明,该算法可以解决大角度斜光照和极度暗光照情况下光照补偿问题,且算法简单计算量小。

An illumination compensation algorithm for human face is presented based on two-dimensional image information.On the assumption that human faces share similar shape with spherical surface,we firstly estimate the face information under uniform illumination along the symmetry axis of face so as to build a standard illumination model by data fitting to the prior statistic information.Then,we analyze the statistical distributions of the grayscale of face image along the direction of light.At last,using the standard model,combined with linear transform and non-linear transform,we can rectify the face image under non-uniform illumination to the standard state.The simulation results on the Yale B face database have shown that the proposed method can realize effectively illumination compensation under the condition of wide-angle oblique light and very dark light.The algorithm has a small computational complexity.