Takuto Shigenobu
ABSTRACT
This paper proposes a three-dimensional (3D) entire shape reconstruction method that performs simultaneous 3D registration of multiple depth images obtained from multiple viewpoints. With the combination of a silhouette-based objective function and evolutionary computation algorithms, the proposed method realizes the entire shape reconstruction from small number (two or three) of depth images, which do not involve enough overlapping regions for other 3D registration methods. In particular, this paper proposes a CMA-ES algorithm with regional intensification techniques (CMA-ESPR+VF) to speed up the registration process. Experimental results show that the proposed CMA-ESPR+VF achieved a speedup that is at most 18 times faster than self-adaptive differential evolution.
- Janez Brest, Sao Greiner, Borko Boskovic, Marjan Mernik, and Viljem Zumer. 2006. Self-adapting control parameters in differential evolution: A comparative study on numerical benchmark problems. IEEE transactions on evolutionary computation 10, 6 (2006), 646--657. Google Scholar
Digital Library
- Nikolaus Hansen. 2006. The CMA evolution strategy: a comparing review. In Towards a new evolutionary computation. Springer, 75--102.Google Scholar
- T. Ushinohama, Y. Sawai, S. Ono, and H. Kawasaki. 2014. Simultaneous Entire Shape Registration of Multiple Depth Images Using Depth Difference and Shape Silhouette. In Proc. 12th Asian Conference on Computer Vision (ACCV'14).Google Scholar
Index Terms
- Silhouette-based three dimensional image registration using CMA-ES with joint scheme of partial restart and variable fixing
Recommendations
Deep deformable registration
Deep network has been utilized in designing a deformable registration framework for end-to-end optimization.Deep deformable registration minimizes a cost that is an upper bound to the sum of squared difference image registration cost.Proposed deep ...
Comments