Performance Acceleration for Video Synthesizing Software Targeted to Sports Training Using Multicore Processor

摘要

Due to fast growth of video camera market, referring videos in sports training has been used as one of the most effective methods for improving their performance because athletes or coaches can analyze their performance objectively. Adopting the advices from coaches, athletes can correct their wrong habits in their performance instantly. Although the videos from multiple cameras in multiple angles include much useful information, it is very hard for them to extract the information from the videos during their limited training time. Introducing an actual example of the training field where multiple cameras are used, this paper is focused on development of new video synthesizing software that generates a single video from the multiple videos, applying a sequence of commands to control the synthesis. This paper also describes an optimization technique to perform the video synthesis fast that applies parallel processing on a multicore processor to invoke multiple threads for video synthesis concurrently. Performance evaluation shows validity of the optimization technique.