This paper introduces a high-speed image processing of embedded systems based on FPGA and DSP collaboration used in coal preparation detection technology. The system uses a modular design, including a real-time image acquisition module, DSP image processing module, and peripheral interface module. The image processing required doing the following arithmetic operations, including image smoothing to enhance, edge detection, region segmentation, gray-scale analysis, pattern recognition. After completion of the above algorithm, the system determines the image feature of coal and gangue, particularly their gray scale values and the center of gravity. By analysis of the characteristics of the image, the system identifies the gangue. The image processing is conducive to the follow-up to discharge refuse treatment.