In recent years, with the continuous development of GPU general computing power, more efficient processing technologies have been for image processing. At present, some image processing algorithms have been transplanted to GPU and have good effect in acceleration. However these algorithms do not make full use of the computing power of each processing unit in a hybrid systems composed by CPU/GPU. Based on GPU programming model and parallel algorithm design, we proposed image collaborative parallel processing model at CPU/GPU heterogeneous environment in this paper. The effectiveness of this model in high resolution grayscale image processing was verified by the image median filtering algorithm, which was based on the computing power of each processing unit in heterogeneous system. The experimental results show that the model performs well in CPU/GPU heterogeneous environment and is easy to execute other image processing algorithms.