Denoising with Bias-Free CNNs

 More information on this concept of removing the bias terms of a neural net trained to denoise images, and then using that as a prior for other tasks.  Removing the bias leads to increased generalization.

The project page for this bias free denoising research is here.

The githib link for bias free denoising is here.

The arXiv page for the paper is here.

You can watch the paper presentation titled 'Robust and Interpretable Blind Image Denoising via Bias-Free Convolutional Neural Networks' from ICLR here.

You can read the associated paper pdf here.


There is a related paper titled 'How do neural networks denoise natural images?' you can fid here.

...

You can use this denoiser model as  a prior for solving linear inverse problems.

The githib associated with this research  is here.


...

Image priors, manifolds, and noisy observations

Visual images lie on a low-dimensional manifold, spanned by various natural deformations. Images on this manifold are approximately equally probable - at least locally. The probability of  being a natural image, , is zero everywhere except for  lying on the manifold.

An observed image, , contaminated with Gaussian noise,  is drawn from an observation density, , which is a Gaussian-blurred version of the image prior. Moreover, the family of observation densities over different noise variances, , forms a Gaussian scale-space representation of the prior, analogous to the temporal evolution of a diffusion process.


Least squares denoising

Given a noisy observation, y=x+z, the least squares estimate of the true signal is the conditional mean of the posterior:


...

Algorithm in a nutshell (see paper for details):

  • Use the denoiser-defined gradient to go uphill in probability
  • On each step, the effective noise decreases, and the effective prior becomes less blurred. The denoiser, and thus the gradient step size, automatically adapts to this.
  • This coarse-to-fine optimization procedure converges to a point on the manifold

Comments

Post a Comment

Popular posts from this blog

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

Image
 Like the Doublemint Twins touting the joys of Doublemint gum, 2 GANs are surely better than 1 GAN.  Especially if we package them together inside of one meta GAN module.  And this is exactly what the CycleGAN architecture does. Have you ever harbored dark secrets of turning a horse into a zebra?  The CycleGAN was developed to do just that. Learn how to turn a horse into a zebra.  And more. Now right away you can notice a difference between the image to image transformation GAN architectures we've been discussing over the last few posts.  Those last few posts described systems that learn from a database of matched input-output image pairs. And if your goal is to turn an edge representation into a nicely filled in continuous tone image, it's easy to build your database of matched input-output image pairs that your GAN system can then learn off of.  Take a continuous tone image (which will be the output of the database pair entry), then run it through an edge detector algorithm. 
Read more

Pix2Pix: a GAN architecture for image to image transformation

Image
 I thought following up yesterday's TraVelGAN post with a Pix2Pix GAN post would be useful to compare what is going on in the 2 architectures.  Two different approaches to the same problem. I stole this Pix2Pix Overview slide below from an excellent deeplearning.ai GAN course (note that they borrowed it from the original paper) because it gives you a good feel for what is going on inside of the Pix2Pix architecture.    Note how the Generator part is very much like an auto-encoder architecture, but rebuilt using the U-Net architecture features (based on skip-connections) that fastai has been discussing in their courses for several years before it became more widely known to the deep learning community at large  (and which originally came from an obscure medical image segmentation paper) . So the Generator in this Pix2Pix GAN is really pretty sophisticated, consisting of a whole image to image auto-encoder network with U-Net skip connections to generate better image quality at highe
Read more

Smart Fabrics

Image
"What makes smart fabrics revolutionary is that they have the ability to do many things that traditional fabrics cannot, including communicate, transform, conduct energy and even grow."  This is a quote from Rebeccah Pailes-Friedman, who is a professor at Pratt with a research focus on developing wearable technology and smart textiles. Enhancing textiles can be viewed through different lens.  Enhancements can be aesthetic, or performance boosting, or durability boosting. 
Read more