Video Summarization Based on the Difference in Statistical Histograms of Fuzzy Clustering Features

The quantity of video footage produced has skyrocketed in the last several years. The enormous increase in video content creates issues with content management. Video and image processing technologies need to receive greater attention to handle the increasing number of videos on the internet and to extract reliable information from them. Video summaries offer concise and streamlined depictions of the content of a video stream. This research aimed to build a methodology for video summarization utilizing the difference in the histogram of video frames using three stages. Frames were elicited from the video, and then SIFT detectors were utilized in the first stage to elicit interest points from each frame. Fuzzy C-means clustering was used in the second stage to collect the interest points. In the third stage, the histograms were made via the number of points for each cluster. The cluster number is represented by the x-axis, and the number of points in each cluster is represented on the y-axis. In order to elicit the summary frames, a value was manually entered, and then a query histogram was built based on these values. The Manhattan metric was used to compare the query histogram with all the histograms constructed in the third step. Experimental results displayed that the mean amount of time for clustering baby video is 3.860 with the proposed method, while the time was 9.670 to cluster the same video using all pixels; also, the precision was 94, the recall was 91, and the F1-score was 93.