A Scalable ADMM Algorithm for Rigid Registration

Sanyal, Rajat and Ahmed, Sk Miraj and Jaiswal, Monika and Chaudhury, Kunal N (2017) A Scalable ADMM Algorithm for Rigid Registration. In: IEEE SIGNAL PROCESSING LETTERS, 24 (10). pp. 1453-1457.

PDF IEEE_Sig_Pro_Let_24-10_1453_2017.pdf - Published Version Restricted to Registered users only Download (609kB) | Request a copy

Abstract

A fundamental problem that comes up in computer vision, image processing, manifold learning, and sensor networks is that of registering multiple point sets using rigid transforms. A standard result in this regard is that the least-square formulation of the registration problem admits a closed-form solution for two point sets. However, since the group of rigid transforms is not convex, solving the least-square optimization for multiple point sets is computationally challenging. It was recently demonstrated that the least-square formulation can be relaxed into a tractable semidefinite program, and that the relaxation is provably tight under certain assumptions. The difficulty is that standard solvers for semidefinite programming (e.g., interior-point solvers) cannot be scaled to handle large-sized problems. In this letter, we propose an iterative solver based on variable splitting and the alternating direction method of multipliers. Since each iteration essentially involves an eigendecomposition, the proposed solver can be scaled to problems that are beyond the reach of interior-point solvers. We present results on simulated and real data to demonstrate the potential of the solver.

Item Type:	Journal Article
Publication:	IEEE SIGNAL PROCESSING LETTERS
Additional Information:	Copy right for this article belongs to the IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC, 445 HOES LANE, PISCATAWAY, NJ 08855-4141 USA
Department/Centre:	Division of Electrical Sciences > Electrical Engineering
Date Deposited:	23 Sep 2017 05:11
Last Modified:	23 Sep 2017 05:11
URI:	http://eprints.iisc.ac.in/id/eprint/57854

Actions (login required)

View Item