Multi-objective optimization for parameter selection andcharacterization of optical flow methods

Abstract

tOptical flow methods are among the most accurate techniques for estimating displacement and velocityfields in a number of applications that range from neuroscience to robotics. The performance of any opticalflow method will naturally depend on the configuration of its parameters, and for different applicationsthere are different trade-offs between the corresponding evaluation criteria (e.g. the accuracy and theprocessing speed of the estimated optical flow). Beyond the standard practice of manual selection ofparameters for a specific application, in this article we propose a framework for automatic parametersetting that allows searching for an approximated Pareto-optimal set of configurations in the wholeparameter space. This final Pareto-front characterizes each specific method, enabling proper methodcomparison and proper parameter selection. Using the proposed methodology and two open benchmarkdatabases, we study two recent variational optical flow methods. The obtained results clearly indicate thatthe method to be selected is application dependent, that in general method comparison and parameterselection should not be done using a single evaluation measure, and that the proposed approach allowsto successfully perform the desired method comparison and parameter selection.