Content Based Video Retrieval—Methods, Techniques and Applications

Abstract

Videos are rich information sources than individual images, they are considered as most influential communication media compared to others. The amount of video data produced and dispensed are growing exponentially day by day with the availability of electronic media such as smart phones, handicams etc. and broadband services at cheaper rates, as well as easy accessibility of those media in the market. Video data storage and access founds its applications in different fields such as digital libraries, video on demand, entertainment etc. and these applications are popular and needs regular access of videos from the libraries. All the above said compound reasons demanded the need of development of efficient video management and retrieval systems which can efficiently retrieve videos similar to the query as well as with a less response time. Video retrieval is made possible by searching of the desired video through a user demanded query. The user inputted query may be in the form of representative keywords or a single image or group of images. The video retrieval systems are classified as text based or content based, according to the query inputted by the user. In a text based video retrieval system query is in the form of representative keywords and the database videos are tagged with appropriate text. An example of concept based search and retrieval system is YouTube. The principal drawback in concept based system is mapping of high level or rich semantics to low level features, which is known as semantic gap. Another drawback in concept based video retrieval systems is intention gap, which denotes gap between query at querying time and intention of the search. Several researchers found content based video retrieval (CBVR) system as solution to the drawbacks of a concept based video retrieval system. The main objective this chapter is to provide comprehensive outlook on content based video retrieval (CBVR) system and its recent developments and a new content based video retrieval system that is going to be developed by feature fusion. The generalized algorithm of CBVR and its individual stages such as keyframe extraction and feature extraction also will be described elaborately. This chapter focuses on a brief overview of CBVR, keyframe extraction, feature extraction and feature fusion.