Spatiotemporal wavelet maximum a posteriori estimation for video denoising

摘要

We examine one way to extend recently proposed wavelet-nbased maximum a posteriori estimation rules for image denoising tonvideo. The proposed approach takes into account both spatial andntemporal dependencies between wavelet coefficients, and is gen-neral enough to incorporate different spherically contoured prior dis-ntributions on noiseless coefficients, as well as different spatiotempo-nral coefficient neighborhoods. Presented extensions of the algorithmnhave reasonable complexity and are suited to vectorized, convolution-nbased implementations.