Novel Efficient Content Based Video Retrieval Method Using Cosine-Haar Hybrid Wavelet Transform with Energy Compaction

Abstract

Advancement in digitalization and networking technologies led to easy generation and storage of multimedia data. With this huge data, it's a current generation requirement to retrieve this multimedia data efficiently and effectively. In last decade video retrieval was based on textual annotations. But there are limitations of subjectivity and intuitiveness to textual annotation. There is a new ray of research area of Content Based Video Retrieval. Transformed visual contents of video represent visual contents in frequency domain. This paper explores the Content Based Video Retrieval using Cosine-Haar Hybrid Wavelet Transform with energy compaction using fractional energy coefficients and four similarity measures alias Euclidian distance, City Block Metric, Sorensen distance and Kulczynski distance for performance evaluation of retrieval system. Among the constituent transforms and Hybrid Wavelet Transform, Hybrid Wavelet Transform performs better with 0.024% energy coefficients using Euclidean Distance, which proves worth of proposed technique for faster and better Content Based Video Retrieval (CBVR).