3D wavelet-based optical flow estimation

Abstract EN

In this paper, a new algorithm for accurate optical flow estimation using discrete wavelet approximation is proposed.

The image sequences are always assumed to be noiseless in the computation of optical flow, since there is always a method that can perform such task.

One of the main application areas of the wavelet transform is that of noise reduction in images.

The basic technique is to transform the noisy input image into a domain, in which the main signal energy is concentrated into as few coefficients as possible, while the noise energy is distributed more uniformly over all coefficients.

The choice of the transform is represent an important tool in optical flow estimation.

In this paper, several algorithms of 1-D, 2-D and 3-D wavelet transforms are adapted for the estimation of optical flow for the first time.