A Cut Method for Cutting and Editing Personal Videos Using ST-patches and Sensor Information

Abstract

We propose a technique for automatically editing personal videos using video and sensor information obtained from the camera. In this study, we used personal videos cut-edited by individual users and their editing history to analyze tendencies in cut scenes (used scenes) and cut-edit points. This analysis revealed that the amount of camerawork is high in used scenes. Based on these findings, the proposed technique uses continuous rank-increase measure (CRIM) and motion correlation (MC) values calculated from space-time patches (ST-patches), which capture change in motion, and acceleration and angular speed obtained from camera sensors to select scenes in which there is a lot of camerawork and subject movement. Editing video in this way based on the user's editing tendencies promotes video editing in accordance with the user's preferences. Experimental results show that automatic video editing by the proposed technique achieves a higher degree of viewer satisfaction than a video editing technique based only on ST-patch features.