High-resolution Shape Reconstruction from Multiple Range Images
High-resolution shape reconstruction from acquired range images is a highly promising technique for various applications, regardless of sensing principle. This task involves two processes: registration and surface reconstruction. Those processes are essentially interrelated problems, and performing registration using only the acquired range images causes errors because of the image degradation involving aliasing and measured noise. Also, the shape surface representation needs to have a continuous form and a high degree of freedom with a compact data set. In addition, local smoothness control is an important issue for surface reconstruction.
Based on these requirements, in this paper, we present an algorithm for estimating the motions and surface jointly. The method realized shape reconstruction based on an implicit surface obtained using an RBF function. We demonstrated the potential of this approach with three kinds of real shapes. The experiments were successfully achieved. The obtained results show that we can see details on the surface that cannot be found in a single range image.
The process and the results of high-resolution shape reconstruction
- High-Resolution Surface Reconstruction based on Multi-level Implicit Surface from Multiple Range Images (2012)
- Yoshihiro Watanabe, Takashi Komuro, Masatoshi Ishikawa: High-resolution Shape Reconstruction from Multiple Range Images based on Simultaneous Estimation of Surface and Motion, The 12th IEEE International Conference on Computer Vision (ICCV2009) (Kyoto, 2009.10.1) / Proceedings, pp. 1787-1794. [PDF]*IEEE