Patch-based learning of adaptive Total Variation parameter maps for blind image denoising
This is an incremental improvement for image denoising applications where noise characteristics are unknown.
The paper tackles blind image denoising with unknown noise distributions (Gaussian or Poisson) by learning spatially adaptive regularization parameter maps for weighted Total Variation using a patch-based neural network approach, showing significant improvement over optimal scalar regularization in numerical results.
We consider a patch-based learning approach defined in terms of neural networks to estimate spatially adaptive regularisation parameter maps for image denoising with weighted Total Variation (TV) and test it to situations when the noise distribution is unknown. As an example, we consider situations where noise could be either Gaussian or Poisson and perform preliminary model selection by a standard binary classification network. Then, we define a patch-based approach where at each image pixel an optimal weighting between TV regularisation and the corresponding data fidelity is learned in a supervised way using reference natural image patches upon optimisation of SSIM and in a sliding window fashion. Extensive numerical results are reported for both noise models, showing significant improvement w.r.t. results obtained by means of optimal scalar regularisation.