Search

Based on OpenGL, this paper presented a local 3D scene reconstruction from the frame buffers (including color buffer and depth buffer) after 3D scene rendered. According to projection principles and depth buffer characteristics, it discussed an inverse prooection processing and obtained theoretically a 3D vertex reconstruction method. As a result, it transformed a l valid pixels within frame buffers and generated a 3D vertex set which was a point cloudy model in the camera space in the conducted experiments. Obviously, this cloudy model was a kind of a sample of the original 3D scene in current camera system and named as virtual vision model(VVM). Secondly, it analyzed the point-wise precision of VVM detailed. The results of experiments show that VVM appears very high relative accuracy to the original 3D scene. Above all, the reconstruction method appears natural multi-resolution feature with respect to the resolution of frame buffers and supports scene simplification and streaming transmission. On the other hand, this method is capable of abstracting VVM real-time from the common 3D software easily without any details of 3D scene known previously, which provides a new approach acquiring 3D data efficiently. [ABSTRACT FROM AUTHOR]