Video shots retrieval with use of pivot points

Abstract

Intelligent analysis of video data is inextricably linked to methods aimed at reducing the amount of initial data necessary for processing in various ways. In this paper, we propose an approach that allows us to reduce the amount of processed video data by excluding it from consideration that is inappropriate for the query. This is achieved by the pivot points analysis of the original data clusters. If the pivot point to be compared is far from the query, then the entire cluster is also far from the query, respectively. Thus, it is possible to significantly reduce the number of operations of query comparison with data and, accordingly, speed up the process.