In order to improve discrimination of the detailed information of dim-target images, the authors propose a new PDE-based nonlinear image enhancement algorithm, which takes the spreading speed of images that vary with the size and direction of the gradient. As a result, the dim-target images can be adaptively enhanced locally, the edges are retained better, and the false edges and the "ladder-effect" are eliminated. Furthermore, the visual effect is improved and processing is more efficient. The experimental simulation results show that the algorithm has higher LSCR and higher GSCR than the traditional methods, and image processing efficiency is greatly increased.