Image Segmentation Using Region-Based Latent Variables and Belief Propagation

We derive a deterministic algorithm that restores and segments an image using belief propagation and a variational Bayesian method based on region-based latent variables and a coupled MRF model. This algorithm estimates two hyperparameters as well as infers the original image and the latent variables. In addition, the algorithm carries out model selection by minimizing the variational free energy. Through experiments using artificial images and a natural image degraded by Gaussian noises, we show that the derived algorithm has the potential ability to restore and segment using a single noisy image.