also MDeGrain being one of the inter-patch (self-similarity) based approaches, is theoretically equivalent to a sparse coding unit, you can have one or many of those sparse coding layers in your denoising autoencoder and it would reasonably give you results similar to MDeGrain but better
|