A Research And Implementation Of Multi-sighted Stitching In Streaming Media Player

Multi-Sighted video stitching can splice several videos, which have overlapping areasand were shot by several cameras, together to a wide viewing angle video or even a360°panoramic video. And now video stitching is a hotspot in research of computer graphics.Multi-sighted stitching streaming media player is a network player which applies videostitching technique, it just need several ordinary cameras shooting a similar perspective ofmulti-channel videos, and the it provides a wide viewing angle or panoramic video.In this thesis, I proposed a real-time video stitching algorithm, based on SIFTalgorithm, to implement the splicing functions of the multi-sighted stitching player. Thealgorithm is divided into two stages: parameters calculation stage, and real-time splicingstage. In the first stage, SIFT features are extracted from video frames to match each other,and in the second stage, video frames are perspective transformed and spliced into onemulti-sighted video frame. Then, a efficiency optimization on real-time splicing stage israised to improve the efficiency of splicing stage. This thesis presents a HSL colorspace-based brightness adjustment solution to eliminate the traces of the video stitchinggap. Finally, a video sequence recognition algorithm, based on SIFT algorithm, is raised,and the algorithm can automatically identify the video sequence.After solving the efficiency of real-time video splicing and display related issues, thisthesis give the system design and the corresponding workflow of the video player.Followed with a comprehensive testing and analysis on the video splicing function of theplayer, the test results proved the system is fully matched with the functional andperformance requirement.