DSP image process with optimized data schedule model

This paper introduces a DSP optimization model based on data structure transformation and memory schedule for real-time image process. This model implements 4 critical methods: image block, image dimension reduction, DMA transfer and ping-pong cache, which makes use of DSP hardware feature and optimizes the data schedule among the external and internal memory and CPU, while processing 2D digital signal such as images and video frames. This model provides a common strategy which can be used in the most of linear image process circumstances. Based on this model, this paper optimized a grayscale image matching algorithm and improved its efficiency, in order to verify the effectiveness of this model.