A Novel Approach to Key Frame Detection Using Histogram and Dynamic Clustering

Abstract

Video analysis is the ability to automatically analyse a video for its temporal and spatial events. Videos consist of frames. Thus, to analyse a video, each and every frame of the video needs to be analysed. High-performance computing is needed for this. Keyframes can be extracted from the videos and used to lessen the computational load. Keyframes are the exemplary frames that contain the significant data needed for analysis. This study suggests a method for identifying the keyframes using the histogram of the frames and dynamic clustering. An average of 97% decrease is obtained in the number of frames required for further analysis.