Video clip retrieval: An integrated approach based on KDM and LBPV

Abstract

Videos are warehouse of information. A good retrieval system is essential for timely access to definitive and accurate information. This paper presents an approach for retrieving shots containing similar text using query by example paradigm. Localization of text is based on Kirsch Direction Mask (KDM). From the localized text region, texture features viz. Local Binary Pattern Variance (LBPV) are obtained. Using LBPV features shots/clips with similar text in a video are retrieved. Experiments conducted on a collection of videos demonstrates that the proposed approach is suitable for video retrieval task.