非线性复合邻域人脸光照补偿

Facial illumination compensation based on nonlinear local remediation

变化的光照是提高人脸正确检测率的瓶颈,提出了非线性复合邻域光照补偿方法来解决人脸检测的光照问题。非线性复合邻域光照补偿方法用对数函数做为基函数,设置的平移系数和分界值系数可以同时对过亮和偏光的图像进行补偿,引入了邻域光照补偿公式解决了传统非线性方法不能针对同一灰度值像素在不同局部应该进行不同补偿的问题。实验结果表明,该方法对过暗和过亮的人脸图像都可以进行有效的光照补偿,熵比较优于传统非线性光照补偿、Gamma校正和直方图归一化。

The variation of illumination is a choke point of imporving the rate of correct face detection.To deal with this problem,this paper proposes an illumination compensation method of nonlinear local illumination remediation.The nonlinear local remediation uses logarithm function as primary function.It meets human visual trait;and sets up two coefficients of offset and one coefficient of Gamma division.Thus this method is able to compensate all types of image through adjusting these coefficients;and introduces a local illumination remediation formula to resolve the nonlinear problem,which uses sole formula to compensate illumination in all local area.Experimental results show that this method is effective in compensating images of over-lighting,over-shading,and partial-lighting,surpasses traditional nonlinear transformation and Gamma correction