Making Use of a Prior Implicit in a Denoiser

 This is a recent talk from Eero Simoncelli of NYU that covers examining and using a prior implicit in a denoiser.  This is really a talk about human perception and perceptual representations of images that uses denoising as an entry into that exploration.

Let's check it out.


1:  Removing the bias in CNN makes it work better (better generalization).

2:  System learns adaptive filters (directional based on input statistics of source and noise).

3:  Projection into a low dimensional space.  Sharp scale space, not blurry.

4:  Spatial derivatives at the high end of the subspace.

5:  Visual images lie on a low-dimensional surface (manifold).

6:  Using a denoiser to do iterative gradient ascent to get to a manifold surface.  This ends up being a generative model of natural images.

7:  Adding a little bit of injected noise to #6 makes it converge better.

8: You can use this 'learned denoiser system' to solve linear inverse problems.  Problems like inpainting, random pixel dropping, super-resolution, random pojections.

Not sure i get that these tasks are called linear?  Linear in a basis set?


The paper titled 'Solving linear inverse problems using the prior implicit in a denoiser' can be found here.


Popular posts from this blog

Simulating the Universe with Machine Learning

CycleGAN: a GAN architecture for learning unpaired image to image transformations

Pix2Pix: a GAN architecture for image to image transformation