Key-Frame based Video Summarization using Optimized K-means Clustering

Video content has been increasing at a very high pace, so summarizing videos is of urgent need. Video summa-rization emphasizes the quick go-through of video content. In the past decades, the field of video summarization has attracted a lot of research. Many of the techniques have been proposed by different researchers. However, most of them are not able to create summaries of general videos, i.e., they are generally good for a set of a category of videos. So, to solve this problem, we have proposed a deep feature-based clustering framework for video summarization. We have used the pre-trained resnet151v2 model for feature extraction, and based on the extracted features, we group the frames of the video using the K-means clustering algorithm. After that, we selected the central frame from each video cluster and finally combined them to form the video summary. The experiments were performed on different types of videos taken from the different datasets and the results show that the proposed methodology performs better compared to some existing video summarization techniques.