Research on Fast Algorithm for Video ROI Detection

It is a hot topic to combine visual perception principles in video encoding. A fast algorithm for region of interest (ROI) detection based on human visual system (HVS) and coding information is proposed in this paper. Low-level coding information is employed as the characteristics of visual perception analysis in order to avoid a large amount of computation for visual perception characteristics analysis processing. Then, video ROI detection is carried out in temporal domain and spatial domain respectively. Last, combine the two domains saliency detection results to mark the video ROI level. The simulation results validate that the proposed algorithm by means of fast motion vector (MV) noise filtering and translational MV checking procedure, then realize fast visual perception characteristics analysis and ROI detection effectively.