Video Skimming for Lecture Video Sequences Using Histogram Based Low Level Features

Abstract

In this article, we create a capsule of a lecture video using visual quality techniques which is a abstract of the lecture video. We first perform temporal segmentation to divide the video into number of shots. The activities of a lecture video can be divided into three categories: slide show, talking head, writing hand. We have developed a different and new approach for determining the quality of non-content and content frames crucial in the lecture video. Different features of the frame histogram are exploited to derive a quality factor for the detected frames. Hence we are able to extract high quality frames with high information content as well as few non content frames (talking head frames) for video clip selection. Finally in recreation of media to produce the capsule of the video, we select a video segments around these selected non content and content frames. Using this technique we are able to obtain a preview of the video which can be viewed in lesser time, hence it is beneficial to the viewers.