PSO-Based Video Summarization for Content Browsing and Visualization Using Parallel Processing

Video abstraction or summarization is a technique that generates a short summary of the video possibly as a group of stationary keyframes. Video abstraction has several applications including video content browsing and visualization. The goal of this research study is to develop a comprehensive approach for generating different sets of keyframes according to the application while speeding up the processing time and minimizing human intervention. The proposed approach relies on dividing the video into equal short segments to avoid the delay resulting from shot segmentation. Keyframes are then generated from these segments using Particle Swarm Optimization (PSO) for additional speedup. Since different runs produce different sets of keyframes, automatic integration of the results of few runs has been shown to produce a set of keyframes capable of capturing the changes within each video shot for video content browsing without the need for human intervention. Since this implies the increase of the overhead, parallel processing is utilized for further significant speed up. The generated keyframes are filtered to produce a satisfactory set of keyframes for video visualization while minimizing the false rate, increasing the hit rate, and minimizing human intervention.