Robust Surface Pattern Matching for Ball Spin Estimation with Multi-exposed Images Under Varying Illumination Conditions

In this paper, a new robust surface pattern matching technique is proposed for estimating the spin rate and spin axis of a spinning ball from a multi-exposed image under varying illumination conditions. In the proposed technique, robust surface pattern extraction based on the OFF cell model of the human visual system is employed to obtain high-contrast surface patterns and a novel rotational surface pattern matching based on the contrast response difference between the patterns is then used to estimate the spin rate and spin axis of the ball with high accuracy. The proposed technique is highly robust and highly accurate when compared to correlation-based techniques. In addition, the processing time for the new technique is less than in the other techniques. The performance of the proposed technique is verified through real experimental results under various illumination conditions.