The Study Of Video Mosaic Method Based On Feature Point

The video mosaic is a technology that splice several scene sequence video which haveoverlapping regions, and form the wide scene or panoramic video which has largerperspective and higher resolution. As a hot research field of video applications, the videomosaic technology is widely used in video surveillance, medical, and other areas.This paper studies the video mosaic it capture from multiple stationary cameras. Byresearching and Comparing image mosaic technology related theory and existing methods，we proposed a video mosaic method which combine phase correlation with the Harrisoperator. First, from the first corresponding frames of two video streams, phase correlationwas used to roughly compute the translation offset between the two streams by each after acertain time interval. Then, in the overlapped region, Harris method was used to detect andregister corners and NCC was used to match the Angular point features. RANSACalgorithm was used to eliminate outliers to ensure effectiveness of the matched corner pairs.Then it computes the homography of each camera, and a large video sequence iscomposited from the multiple video sources.In this paper, selecting the feature points detection operator go through a lot of analysisand comparison. At first, focus on the corner detection principle and process of the Moravecoperator, the SUSAN operator and Harris operator. By a wide range of analysis, selectingthe Harris corner detection operator which has the better performance. Because of thereal-time requirements of video mosaic process, this paper advances a method that combinephase correlation with the Harris operator. The main idea is reduce the detection area ofcorner. Detecting and matching the angular point in the overlapping area could greatlyreducing the complexity of the algorithm, improving the computational speed and meetingthe real-time requirements of video mosaic. Finally, experimental results demonstrate theeffectiveness of the algorithm.