Harmonic Analysis and Signal Processing Seminar
Motion-based 3-D wavelet frames and probability models for video processing
Ivan Selesnick, Polytechnic
University
Wednesday, September 29, 2004, 2-3:00pm, WWH 1314
Abstract
The denoising of video data should take into account
both temporal and spatial dimensions, however, separable 3-D transforms have
artifacts that degrade their performance in applications. True 3-D transforms
are rarely used for video denoising. We describe the design and application
of the non-separable 3-D dual-tree complex wavelet transform for video denoising.
We show that this expansive transform gives a motion-based multi-scale decomposition
for video - it isolates in its subbands motion along different directions.
The development of this transform depends on the design of pairs of wavelet
bases where the wavelet associated with the second basis is the Hilbert transform
of the wavelet associated with the first basis. We also illustrate the modeling
of the wavelet coefficients as mixtures of Laplacian random variables.