Efficient implementation of optical flow algorithm based on directional filters on a GPU using CUDA

This paper describes an optical flow estimation algorithm using directional filters and an AM-FM demodulation algorithm, and its efficient implementation on a NVIDIA GPU using CUDA. The resulting implementation is several thousand times faster than the corresponding MATLAB code, which makes the described scheme suitable for real-time applications. This paper also describes a new multiresolution scheme for our algorithm which allows estimating high speeds without aliasing. The accuracy of our algorithm has proved to be comparable to the accuracy of well-known optical flow algorithms.