Interactive 3D Modeling And Depth Recovery From Multiple Images

With the rapid development of computer vision, virtual reality and multimedia technology,3D models have been widely used in various fields, the problem of how to produce 3D models receives unprecedented attention. Currently there are three main ways to reconstruct 3D models. The first is the traditional interactive modeling based on professional software, which is technology ripening but require extensive manual interaction. The second is using 3D laser scanner to acquire high-precision models, it is very convenient but can not acquire the texture of models in most cases. The last way is 3D reconstruction based on multiple images, which can generate both of models and texture. However, the robustness is still an issue because of many factors, such as noise, highlight, reflection, and occlusion. Fortunately, we can solve these problems properly with few manual interaction.In this paper we present an interactive 3D modeling system based on image sequence. We propose two different interactive approach named plane-based and axis-based methods, with the observation that many objects have regular geometry structure in our real world. Besides extracting 3D models, we apply our modeling system to recover the depth of video sequence. Compare with the original interactive depth recovery system, our system not only improve the depth quality but also enhance efficiency of the interaction.