Setup. We consider a noisy observation model , where:
is a 2D spatial coordinate,
is the true image,
is a white Gaussian noise realization with variance .
Objective. Our aim is to produce an estimate of the true image , given a noisy observation and the standard deviation of the noise, .
Approach. We propose a novel approach which is based on effective filtering in 3D transform domain by combining sliding-window transform processing with block-matching.