Range Image Flow using High-Order Polynomial Expansion

Abstract : This paper presents a novel algorithm for estimating the motion in a series of range images. First, each range image is approximated by applying a high-order polynomial expansion to local neighborhoods within the range image. Then, these approximations are used to derive the translation or displacement estimation from frame to frame within the series of range images (also known as range image flow). An iterative method for computing the translation is presented. We evaluate the algorithm on several synthetic and real-world range image sequences with promising results. Results in this paper are generated from a single iteration of the algorithm in space and time. Therefore, our next step is to improve the implementation of the algorithm so that using multiple spatial scales and past information improve the signal flow estimation, as we would expect. Also, we will port the MATLAB implementation to C++ to improve speed and meet the signal integration goal of the algorithm.