Probabilistic image modeling via neural networks

The authors propose a neural network approach for the estimation of the local conditional distributions of textured images. They use these distributions to generate a probability distribution on the entire image. The proposed approach overcomes many of the difficulties encountered when using Markov random field (MRF) approaches. In particular, the approach does not require the trial-and-error choice of clique functions or the subsequent estimation of clique parameters. Simulation results show that the images synthesized using neural network modeling produced desired textures more consistently than MRF/Gibbs distribution based methods.<>