An Optimization of Video Sequence Stitching Method
Hong Yu, Liuyang Kong · 2018
Large field of view with high resolution has always been sought after for security monitoring personnel. Although the video sequence stitching has already been developed for several years, most existing method cannot achieve real-time stitching and fail to get excellent results. In this work, We used the same set of video frames to compare the time and the detection rate of three different feature point detection algorithms: SURF, Harris and ORB. In addition, we proposed a novel image fusion method. We calculated the region boundary and did morphological erosion operation. The pixels that need to be fused are determined, and the trigonometric functions are combined as the weights to be fused. Experiment shows that our optimized method not only achieve real-time stitching, but also meet the high quality requirements.