Performance comparison of content based and ISODATA clustering based on news video anchorperson detection

Abstract

In this world, large amount of information is stored in our database. These large quantity of information are stored in different places in our database. Therefore improvement has to be made in extracting essential information from the database. Various algorithms are available in anchorperson detection. Among those content based method is existing in anchorperson detection. Although it does not effectively detect the anchorperson frame where some anchorperson frames are missed in this method. The anchorperson frames are correctly detected and missed frames are detected by using ISODATA clustering algorithm which is proposed in this paper. In this method the amount of data in video is efficiently reduced by using visual keyframe descriptors and clustering algorithm. After that the existing and proposed algorithms are compared by using precision and recall rates. The experimental results show that the ISODATA [Iterative Self Organizing Data Analysis Techniques Algorithm] clustering can cluster the video and the method is efficient and gives a robust performance.