Models of multiple inference in statistical fluctuation for image restoration

Models of multiple inference in statistical fluctuation are presented for restoring a degraded image. Multiple images inferred by the Potts model are prepared in the initial stage. By using the inferred images, two approaches are described. The first approach is that a pixel value of a restored image is a median value of inferred images for respective pixels. The second approach is that a pixel value is an average value of them. The effectiveness of our approaches is confirmed through numerical experiments.