In this fast-paced world, it is a tedious process to spend time separately for reading through a number of slides. To the date various researches have been done in order to shorten the preview content such as summarizing spoken lectures by combining lecture materials and by using automatic indexing systems. The solution proposed for the problem involves applying computer vision techniques like background subtraction and inter-frame difference on the input slides and picking out the high ranked slides. Also, the textual content present in the selected slides are further summarized to provide faster understanding. The novelty in this work considers the foreground image that is obtained by subtracting the background image and inter-frame difference and calculates the pixel-based score slide. Using this pixel-based score, assigns the score for every slide and ranks them. Then select the top-ranked slides for summarization. The performance of the proposed model is to be evaluated by providing shorter versions of the lecture slides to students and evaluating their performance in a questionnaire conducted afterwards. Another novelty in this work is that, processing of slides is done in parallel, thereby, reducing the computation latency and cost.
Volume 11 | 02-Special Issue
Pages: 1885-1894