Shot Segmentation And Video Retrieval Based On Multiple Features

Abstract

Rapid growth in the field of data capturing and storage along with newlinethe increase in the availability of multimedia content in the web has resulted newlinein many large personal and public digital video databases However the rapid newlinedevelopment in the availability of the multimedia database is not newlineaccompanied by the technologies used for its efficient usage and retrieval newlineThis is due to the fact that the content of multimedia database is not same as newlinethat of the text database which is easily accessible based on the keywords of newlinethe document Therefore it is very important to segment and organize the newlinevideo data so as to be accessed easily It induces the interest of researchers to newlinefocus more on the area of video processing Although many research works newlinehave been carried out and many systems have been developed in the area of newlinevideo segmentation and video retrieval both computational cost and accuracy newlineof the existing systems are still far beyond users satisfaction newlineThe main issue to be addressed in the case of video segmentation is newlinefeature selection which is robust to illumination camera and object motion newlineand a measure of dissimilarity to detect the boundaries In this thesis an newlineattempt has been made to address the issues involved in shot segmentation newlineand video retrieval which are the two important tasks of video processing newlineThe aim of this thesis work is to improve the performance of the Shot newlineBoundary detection SBD and Content Based Video Retrieval CBVR by newlineproposing new multiple and effective combination of features newlineAn enhancement in the existing Localized Edge Block LEB newlinemethod is proposed in view of improving the performance by overcoming the newlinedrawbacks of using intensity based features in the existing LEB method and it newlineis named as Enhanced Local Edge Block ELEB In the first proposed newlinesystem the feature based on phase angle variation of the edge pixels is newlineproposed to get more detection accuracy and eliminate camera and object newlinemotion In the existing LEB system cut and dissolve detections are carried newlineout separately newline newline