A pyramid model for blind deconvolution based on simulated annealing
Yen‐Wei Chen, T. Enokura, N. Mendoza, Zensho Nakao · 2002
A pyramid model is proposed for blind deconvolution problem of image restoration based on simulated annealing (or other heuristic method). The proposed approach is like a pyramid. The authors start the restoration from a low resolution level. As the image with the low resolution is restored, the restored image (low resolution) is expanded to an image with higher resolution, which is used as an initial estimate of the high-resolution level. The restoration process is repeated to the highest resolution level. Since one can start with an initial estimate, which will be near the optimum solutions for each resolution level even for the highest resolution level, one can easily find the optimum solutions (restorations) with high resolution. Computer simulations have demonstrated the applicability of the proposed method.