A Direct Geometric Algebra Optical Flow Solution Abstract

Optical flow is the vector inverse problem of estimation motion though an image sequence. Geometric Algebra is an appropriate mathematical language for describing and solving vector problems. In this paper we apply Geometric Algebra to optical flow and pose a direct solution using a simple smoothness constraint. Implementational considerations are given and comparative assessment is made to two seminal optical flow methods, namely the Horn-Schunck and Lucas-Kanade algorithms. The new method is demonstrated to be fast and to give results comparable to these seminal methods