Improving Clustering Efficiency with MSER and OCR Representation in Video-Theme Detection

Abstract

In present days, the more authorized news was located in different websites or journals in different languages. The videos may be in many formats such as avi, mp3, etc. In this paper, we proposed an efficient method for automatically detecting and tracking the topics from the videos by using Maximally Stable Extremal Region (MSER) algorithm which is mainly used for blob detection. The news topics are detected by using clustering sampling process. For clustering we use an efficient algorithm Optical Character Recognition (OCR) to get an optimal clustering solution. The outcome of the proposed project will prove that textual and visual representation of videos produces the appropriate topic by using AND-OR Graph Representation. This method achieves the maximum efficiency and higher performance in the clustering process.