A spatio-temporal filtering method for motion estimation
Rui Wang, Yan Fei Zhao, Yujun Tang, Yan Yuan · 2011
Motion estimation (ME) from image sequences is of primary importance for applications like target tracking, velocity measurement and vision guidance. By the fact that the wavelet transform has the ability to handle the signals with the non-stationary, time-varying properties, a novel spatio-temporal filtering approach based on the Continuous Wavelet Transform (CWT) to realize the motion selectivity is presented. The Fourier analysis of the signal in one dimension of space and time as well as the wavenumber-frequency domain coverage of the Morlet wavelet for different parameters are adopted to provide the strategy for motion estimation and tracking target which moves in a relatively simple mode. Experiments take place with real video sequences to demonstrate that our method has the advantage of being robust to illumination variations, noise, and being computationally efficient.