Image enhancement using very fast simulated reannealing

We compare the random greedy algorithm, the Cauchy annealing algorithm and the very fast simulated reannealing algorithm (VFSR) on a piecewise constant image enhancement task. We tested these algorithms on a noise corrupted 64/spl times/64 and 256/spl times/256 image and show that the VFSR algorithm converges to better solutions than Cauchy annealing, or the random greedy search algorithm.