Consistent labeling with PDP models: benchmarking studies

Summary form only given. Relaxation labeling (RL) processes and Gibbs sampler (GS) can be considered a class of iterative parallel algorithms that solve the consistent labeling problems. They are widely used in image processing and the recognition of figures by way of reducing ambiguities of labeling. Although they have similar properties, their iterative improvement methods are quite distinct. That is, RL is essentially a deterministic process while GS is stochastic. The author has evaluated and compared the performance of the models for coloring problems and found from the experimental results that for this kind of problem RL is more efficient than GS.<>