Loss functions for image processing?

Does anyone know of a good place to go to see the different loss functions and explain how each one is useful for image processing? I am building a upressor and a deconvolve and I am having problems with lots of artifacts. I was wondering if there are other loss functions I can mix in there to help optimize for a smooth image without a lot of noise.

You can check the total variation loss (here).
It was introduced in Neural Style Transfer exactly to avoid noisy artifacts.

Awesome, thank you! I am used to just working within fastai so this will be a good test to see if I can implement something from regular pytorch into fastai.