Parameter estimation of mean field annealing technique

We propose a method of parameter estimation using order-of-magnitude analysis for optimal boundary smoothing in the mean field annealing (MFA) technique. The MFA method smooths out the noise without losing the sharpness of the corners. However, the MFA algorithm is influenced by several parameters such as the standard deviation of the noise, the relative magnitude of prior term, initial temperature and final temperature. We propose a general parameter estimation method for optimal boundary smoothing using order-of-magnitude analysis for consistent image representation. In addition, we prove the effectiveness of our parameter estimation and also show the temperature parameter sensitivities of the algorithm.