Discrete wavelet analysis: A new framework for fast optic flow computation

Optical flow estimation in omnidirectional images using wavelet approach

“…Two sequences (fig 1, fig 2) where the camera undergoes an uniform translation motion and two sequences (fig 3, fig 4) where the robot turns on itself. In each case we have to compare our results with the method proposed by Bernard [2]. This method is also based on a projection of the Brightness Change Constraint Equation (1) on a basis wavelets but it supposes the flow is locally constant on the support of this one.…”

“…and the approximation [2] show that if we replace is not large, i.e there exists M such that v < M s…”

“…When we deal with planar images, it is usual to assume that the flow is locally constant [2], [8]. With the omnidirectional images, this hypothesis is never valid.…”

“…To fulfill the requirement of real times application, we follow the method proposed by Bernard [2] which carries out a good compromise between precision of optical flow estimation and speed of computation. We will show this method requires a multiresolution approach to compute the optical flow.…”

Motion Estimation in the 3-D Gabor Domain

“…In [30], an adaptive method is designed to choose the best scales to avoid the aperture problem in estimating the local motion field. In [31], a successive approximation scheme is proposed to compute optical flow in a Laplacian pyramid structure.…”