Keyframe Extraction for Low-Motion Video Summarization Using K-Means Clustering

The rate of increase in multimedia data required the need for an improved bandwidth utilization and storage capacity. However, low-motion videos come with a large number of feature-related frames due to its static background. These redundant frames result to difficulty in terms of video streaming, retrieval, and transmission. In other to improve the user experience, video summarization technologies were proposed.  These techniques were presented to select representative frames from a full-length video and remove the duplicated ones. Though, an improvement was recorded in the keyframe extraction process. However, a large number of redundant frames were observed to be extracted as keyframes. Therefore, this study presents an improved keyframe extraction scheme for low-motion video summarization. The proposed scheme utilizes a k-means clustering approach to group the feature-related frames within a given video data into number of clusters. Furthermore, a representative frame from each cluster was extracted as keyframe. The results obtained shown that  the proposed scheme outperforms the existing scheme in terms of compression ratio, precision and recall rates with a value of 26.62%, 13.78%, and 6.63% respectively