A Novel Hybrid Keypoint Detection Algorithm for Gradual Shot Boundary Detection

In the recent period, video summarization plays a vital role in the video retrieval process especially in the video surveillance system. The summarization process comprises of several activities such as preprocessing, features extraction, frame similarity analysis, shot boundary detection (SBD), etc. This paper presents a novel hybrid keypoint detection algorithm for the gradual shot boundary detection. The proposed technique combines the Hessian matrix of point of interest and the minimum Eigen values of the region of interest. The proposed technique has been tested on several video data sets and the results shows better performance in terms of accuracy when compared to the previous methods such as MEV, CD, FAST, SURF, MSER and BRISK techniques.