A Nonrigid Image Registration Approach Based on Cooperative Game Among Pixels

Abstract

After represent the deformation as a field of discrete displacements, nonrigid image registration problem can be considered as a maximum a posteriori inference problem in Markov random field, which can be optimized using pairwise Gibbs energy minimization technique. To find a minimum, we model the registration problem as a game among pixels, propose a novel approach for nonrigid image registration based on cooperative game theory, and design a deterministic optimization algorithm. During the game, each pixel tries to form a coalition with its neighbors, and choose a displacement to maximize its own payoff and the coalitional payoff. Registration is achieved by an iterative process. The novel approach can register any complicated deformation robust. Experimental results show that the cooperative game-theoretic approach for nonrigid image registration is feasible and effective.