• Setup. We consider a noisy observation model MATH, where:

    • MATH is a 2D spatial coordinate,

    • $y$ is the true image,

    • MATH is a white Gaussian noise realization with variance $\sigma ^{2}$.


  • Objective. Our aim is to produce an estimate $\widehat{y}$ of the true image $y$, given a noisy observation $z$ and the standard deviation of the noise, $\sigma $.


  • 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.