Velocity measurement using phase fitting of analytic spatiotemporal images

In this paper, the problem of one-dimensional (1-D) velocity estimation is addressed, using two-dimensional (2-D) spatiotemporal orientation estimation. A new frequency estimator based on a least square plane fitting of the estimated autocorrelation phase has been previously developed and is applied here to random textured images for the problem of velocity estimation. This algorithm requires the computation of the analytic autocorrelation and of a 2-D phase unwrapping step. Then, a plane fitting applied to the unwrapped phase gives an estimation of the spatiotemporal image orientation. Finally, we show in this work that the velocity can be estimated successfully in the spatiotemporal plane with the use of a 2-D frequency estimator.