Binary Image Denoising | Iterated Conditional Modes | Simulated Annealing | MAP | MRF | python
Binary Image Denoising with greedy ICM (Iterated Conditional Modes) algorithm and global optimization algorithm SA (Simulated Annealing) - Bayesian MAP inference in (discrete) Markov Random Field UGM (undirected graphical model) with Ising Priors, an implementation in python
ICM is greedy, fast, always rejects locally bad update and converges to local minimum often
SA has metropolis criteria where it can accept/reject an update, initially ar high temperature accepts lots of bad updates to explore the space (escape from local minimum) and gradually cools down, helping to converge to a better (sometimes global) minimum.
Pixels are visited in random order
#imageprocessing #imageprocessingpython #python #machinelearning #optimization #algorithm