Computing three-dimensional motion parameters: a hypothesis testing approach

摘要

Despite great advances in the analysis of motion problems, implementing or searching for correct and robust algorithms is still challenging and elusive. The main purpose of this paper is to show that a hypothesis testing paradigm can provide an effective and robust way for computing solutions to motion problems. In particular, we show that computing a solution to a motion problem is equivalent to selecting a correct hypothesis from a set of possible ones, and rejecting or validating a hypothesis involves similar computations. The proposed method is inherently suitable for implementations in parallel machines. We use a variety of real images, like outdoors scenes, truck motion, and robot scenes to test and evaluate our method.