An image enhancement model based on the Human Visual System is proposed. Image enhancement processing normally decreases image quality in some aspects, because the representation of color space is not linear.The Retinex algorithm can be generally used in mamy kinds of images by considering human visual perception. The existing Retinex algorithm both Single-Scale Retinex and Multiple-Scale Retinex, do not take into account the local variations of the contrast within an image scence. Therefore the performance advantage of these models is limited.A modified Retinex algorithm by using non-isotropic Gaussian kernel filters to fit the human visual system and the image perception quality is increased by equalizing the obtained Retinex ratio in logarithm. The processing is accomplished by cope the various scales of the window support regions yielding a multiscale enhanced images.A number of experiments are conducted to demonstrate the performance of the proposed algorithm.