Abstract: In the remote sensing imaging instant, the uneven atmosphere and illumination often cause inhomogeneous luminance distribution in the images. This nonuniform phenomenon will affect the further applications of these images. In this paper, a new uneven illumination correction method for optical remote sensing image is presented. The method is based on the Retinex theory and the variational function is used to estimate the uneven illumination distribution in the imaging instant. Retinex theory addresses the problem of separating the illumination from the reflectance in a given image, which in general is an ill-posed problem. The color sensation for any area in an image is not dependent on illumination but on reflectance which should be retained. In the variational Retinex framework, the projected normal steepest descent optimization method is applied to solve the function and the multi-resolution numerical solution is introduced to improve the algorithm efficiency. The proposed algorithm was tested on a synthetic image and two real aerial images. Experimental results validated that the proposed algorithm outperforms the traditional methods in terms of the calculation efficiency, the quantitative measurements and visual evaluation.