How do we get a P? Are there distributions that are universally common/applicable or do we have to generate one for each location?
It seems like because the goal is to have p be close to the actual integral function, it would vary a lot between different images. However, maybe there are standard PDFs that are used in different settings - outdoors, inside a closed room with a ceiling light, in a stadium, etc.
I wonder if we can do other sampling methods like Gibbs sampling or Metropolis Hastings for this.
Seems like a lot of similar variables can be defined and used: https://ds-102.github.io/fa19/assets/notes/notes10.pdf