Efficient algorithms for robust feature matching

One of the basic building blocks in any point-based registration scheme involves matching feature points that are extracted from the sensed image to their counterparts in the reference image. This leads to the fundamental problem of point matching: given two sets of points, find the affine transformation that transforms one point set so that its distance from the other point set is minimized. Because of measurement errors and the presence of outlying data points, it is important that the distance measure between two point sets be robust to these effects. We measure distances using the generalized Hausdorff distance. Point matching can be a computationally intensive task, and there have been a number of algorithms and approaches proposed for solving this problem both theoretical and applied. We present two approaches to the point matching problem, in an attempt to reduce the computational complexity of the problem, while still providing guarantees on the quality of the final match. Our first method is an approximation algorithm, which is loosely based on a branch-and-bound approach due to Huttenlocher and Rucklidge. We show that by varying the approximation error bounds, it is possible to achieve a tradeoff between the quality of the match and the running time of the algorithm. Our second method involves a Monte Carlo method for accelerating the search process used in the first algorithm. With high probability this method succeeds in finding an approximately optimal match. We establish the efficiency of our approaches empirically.