Objective: The main motivation of this research is to discover and segment out common object regions in different videos.
Methods: The proposed system introduced a spatio-temporal scale-invariant feature transforms (SIFT) flow descriptor which is used to incorporate across-video correspondence. In order to improve the system performance particle swarm optimization (PSO) is used which captures the optimal inter-frame motion based on the position and velocity updation of the particle. In this optimization process, we use a spatio-temporal SIFT flow that integrates optical flow, which captures inter-frame motion, and conventional SIFT flow, which captures across-videos correspondence information. This novel spatio-temporal SIFT flow generates reliable estimations of common foregrounds over the entire video data set.
Findings: The experimental results show that the proposed system achieves better performance compared with existing system in terms of accuracy, precision, recall and f-measure.
Improvement: The proposed algorithm increases the overall system performances by spatio-temporal scale-invariant feature transform flow descriptor and particle swarm optimization algorithm prominently.
Keywords
