In order to achieve real-time adjustment of the camera and produce satisfactory images, an embedded system oriented approach, which involves DSP and FPGA is proposed. The new method, combining traditional methods of gray scale mean value and information entropy of images has been empirically proved to maximize the information carried by the image, and ensures a proper brightness of outputs, it also quantifies pre-processing with FPGA, and maintains real-time productivity of algorithm with the help of system software in DSP. As is indicated in the experiment with the target system, images under this method are obviously distinct.