Modeling 3D Objects from Range Maps and Color Images using a Warping-based Approach

Two important means of acquiring 3D models of real world scenes and objects are range scanners and stereovision systems. The first approach is usua lly implemented through laser-range mapping techniques, which result in dense depth maps, with relatively high accuracy. On the other hand, stereo methods reconstruct the scene geometry by finding corresponding points in two or more images, then triangulating these matches. Stereovision is one of the earliest methods for the recovery of scene structure from imagery. Unfortunately, building accurate dense depth maps using the latter approach is made difficult by the correspondence problem. Nevertheless, stereovision is still a popular method for many applications. One of its advantages is the possibility of real time implementation in video-based modeling pipelines [1]. In fact, despite the increase in the acquisition speed of high-resolution laser scanners, real time systems are still rare and expensive.