Residual-Consensus Driven Linear Matching

Linear matching (LM) is a simple and effective method for solving image matching problems. In many cases, image matching problems are nonlinear due to involvement of the geometric transformations; therefore, an essential step for utilizing linear models for image matching is to linearize the geometric transformation matrices that introduce nonlinear terms into image matching problems. Existing LM methods usually use low-order transformations to artificially initialize a linear model. In this paper, we propose a residual-consensus driven LM algorithm that generalizes existing LM methods by allowing higher order transformations to automatically initialize a linear model. Based on the observation that transformation models generated from inlier subsets exhibit correlated behaviors (termed residual consensus hereafter), we develop a residual-consensus robust estimation algorithm to project the nontrivial linear transformation problem into a much smaller subspace, and thus enable efficient optimizations through linear programming. The experimental results on synthetic and real databases demonstrate the effectiveness and robustness of the proposed algorithm.