Before And After The Blackfin Dsp-based Video Processing Algorithm Design And Optimization

Video pre-processing and post-processing are widely used and the computational load of them is considerably high. In this thesis, we discuss three kinds of video pre-processing and post-processing algorithm, including video stabilization, deblocking filter and image scaling. Also, we optimize these algorithms based on ADSP-BF561.Firstly, we present the background of video technology and the characteristics of the BlackfinADSP-BF561 DSP.Secondly, our discussion is focus on design and optimizing of the three video pre-processing and post-processing algorithm. We use video pixel instructions of Blackfin DSP to optimize the video stabilization algorithm, and most of the condition branches of this algorithm. The deblocking algorithm is based on MPEG-4 deblocking filter. We use vector instructions and video instructions to write the assembly code. Additionally, we optimize the data flow of this algorithm. Image scaling algorithm is based on linear interpolation. We optimize this algorithm by increasing the cache hit rate, improving the code parallelism and using the circle buffer.The results of experiments are presented after discussing each of the algorithms, and we analyse the results in addition.