IFrames: A Multi-level Keyframe Extraction and Navigation Tool for Videos

Abstract

A video sequence can be represented as a sequence of points in a high dimensional feature space that forms a poly line curve. Mapping a video frame onto a point is based on the geometric features detected in the frame which provide the perceptual differences between frames. The poly line is then analyzed and simplified using two metrics considering the mutual relationships among the nodes. The poly line is recursively simplified and a multi-level representation is obtained. The nodes at the low level poly line correspond to the key frames of the video. After transforming the poly line from the high dimensional space into 3D space, users can view the key frames as well as their mutual relationships in an intuitive and easy manner. The multi-level representation of poly lines offers a coarse-to-fine key frames for the video, and provides users a favorable experience on navigating the video.