The non-uniform distribution of light illumination makes images show some shadow areas and highlight areas in the process of collecting images, which will lose a lot of image details. To solve this problem, this paper proposes a partial differential image enhancement method based on histogram equalization. First, this method transforms the image into a gradient domain, and the new gradient function is obtained through the gradient field transformation. Then, by using the least square principle to reconstruct the image from the gradient field, the textures and details of the image are enhanced and the unknown details are shown all. According to the non-uniform properties of light illumination, anti-balanced transformation is proposed by combining with the histogram equalization to adjust the brightness and contrast of the images. Furthermore, it uses the finite difference method to discrete the image. By combining with the heat equation, the method has characteristics of simple calculation, fast operation, stronger flexibility and wider applications. Based on programming by Visiual C++, its image processing time is 35 ms for an image with the size of 512 pixel×512 pixel, which meets the requirements of real-time image enhancement systems.