Study Of Optimization Strategy For Motion Estimation Algorithm And The System Based On WMSN

Wireless Multimedia Sensor Network(WMSN) is produced by adding such detailed and vivid information as voice, image and video in Wireless Sensor Networks(WSN). WMSN is worthy of extensive research and has a huge market value. A multi node wireless real-time video playback system(call system or the system in short later) researched in this paper belongs to WMSN. This paper studies the optimization of motion estimation in video compression and further optimize to the whole system after embedding the optimized algorithm to the system.This paper firstly introduces the development of WMSN technology and analyzes the system design. Then the system architecture design: The hardware uses TMS320DM642+ PXA270 as the core, software selects H.264 standard in video compression. Finally, in the system slow coding speed and poor play quality problems, optimize the system from the motion estimation and the whole system in two directions.In the direction of motion estimation, this paper firstly introduces the key technology of video compression and motion estimation based on H.264 standard, proposes a custom motion estimation standard for WMSN. Then analysis the existing motion estimation based on block matching method and points out the direction of optimization. Based on the integral image, this paper proposes a fast algorithm to calculate the image of(sub) macro block sum. Then two kinds of optimization strategy for the existing motion estimation algorithm are put forward. One kind of optimization based on the old motion estimation standard, calculates by filtering and simplification, improves the compression speed without affecting the compression result. The other kind of optimization based on the custom motion estimation standard introduced in this paper, improves compression speed with little influence on the compression quality. Simulation and analysis were carried out after the optimization above.In the direction of the whole system, this paper chooses the best one from motion estimation optimization algorithms above and embeds it to the system, then optimizes the whole system as following. Firstly proposes an optimization mode. Then designs, optimizes every module of the model and implements the whole system. Finally analyzes and solves the problems of long time running in the system. The on-line test results are satisfactory, this shows that the research done in this paper is effective, and the system is ready for commercialization.