Visual tempo analysis for MTV-style home video authoring

Abstract

Automatic visual tempo analysis and adaptive shot detection are studied for MTV-style home video authoring. Based upon the significant variation caused by object and background independently, spatial and temporal analysis are integrated for region of interest(ROI) detection in video sequences. To further detect homogeneous shots, visual tempos are proposed to describe the variation. Then, an adaptive segmentation method is presented to automatically detect homogeneous shots using visual tempo cues. A baby video is used as an illustrative example, and the experimental results are encouraged.