1. Introduction
Camera motion classification is an important issue in content-based video retrieval. Extracting camera motion helps understand higher-level semantic content, especially in some specific domains, such as sports video, surveillance system etc. Zoom operation is a basic camera motion that results from the change of focal length, or camera translation along optical axis. Usually, zoom-in operation gives details about the characters or objects do, and imply the important event may happen. Zoom-out operation gives a distant framing and shows the spatial relations among the important figures, objects, and setting in a scene. So zoom identification is essential to the study of video structure and extraction of higher semantic information.