A Weighted Total Variational Method For Retinex

Human visual system can identify the constant color of a scene under varying illumination conditions which is called the color constancy theory.Retinex theory explains this phenomenon and shows that color sensations correlate with the intrinsic reflectance of an object and are independent of the illumination conditions.Based on the retinex theory,the pixel value of the image is determined by two factors.The two factors are the illumination function and the reflection function.The goal of retinex is to decompose the reflectance from a given image taken under a certain illumination and thereby obtaining the real color of the scene.The retinex theory is widely applied in image enhancement,high dynamic range image tonal mapping,satellite image correction,target tracking,image shadow elimination and so on.In this work,we propose a weighted total variational model for retinex problem.An alternating minimization algorithm is devised to solve the proposed model.The experimental results show that the reflectance and enhanced images obtained by weighted total variational model are much better in terms of both preservation.