1. Introduction
The rapid development and reduced cost of digital video capturing devices have made it economically feasible to deploy driving video recorders (DVRs), which can record the scenes as well as events along the way. When a traffic accident has occurred, the police may need to collect videos captured by DVRs on the cars which have passed by the accident scene for investigation. It is laborious and inefficient to watch such substantial video collections in a sequential way for searching relevant video clips. Therefore, we are motivated to develop a street view image generation system for providing a visual summary of street scenes along the way and enabling users to extract a video clip corresponding to a specific road section quickly. On the other hand, the proposed framework can also be extended to many practical applications, such as city profiling, virtual touring, street navigation, introductory building/store indexing, and personal driving/travel recording.