Autoensum: Automated enhanced summary for multiple interacting objects

Abstract

Video Summarization is a promising approach towards concatenating the moving patterns of objects into a single image. The summary attracts readers because of less browsing time, minimized spatio-temporal redundancy, and a feel of motion activity of the scene. It becomes crucial when video uncovers multiple interacting objects and the quality of the video summary in the form of resolution deteriorates in this situation. This paper attempts to address these type of concerns and presents an approach that helps the viewer to have a more automated super resolved summary of the general content of the video. We propose a method, that provides fully automated reference frame selection, frame removal, super resolution and denoising for a productive video summary involving multiple interacting objects. We have evaluated our approach on different types of videos for the purpose of their quantitative and qualitative comparison.