In the literature of computer vision and image processing, motion estimation and image registration problems are usually formulated as parametric fitting problems. Least Squares techniques have been extensively used to solve them, since they provide an elegant, fast and accurate way of finding the best parameters that fit the data. Nevertheless, it is well known that least squares estimators are vulnerable to the presence of outliers. Robust techniques have been developed in order to cope with the presence of them in the data set. In this paper some of the most popular robust techniques for motion estimation problems are reviewed and compared. Experiments with synthetic image sequences have been done in order to test the accuracy and the robustness of the methods studied.
展开▼