Genetic algorithm-based reconstruction of old films corrupted by scratches and blotches

摘要

This paper presents a method for restoring old films corrupted by scratches and blotches. In the proposed method, the reconstruction problem is formulated in a Bayesian framework, where a priori and likelihood are used to model the smoothness within the original images and between the original and observed images, respectively. As the proposed model uses no prior knowledge of scratches and blotches, it can be used to restore various types of degradation. In addition, to minimize the energy function derived in the proposed model, a modified version of DGAs is used, where chromosomes are evolved according to regionally assigned properties, thereby improving the convergence to stable solutions. The proposed restoration method was tested using well-known old films and artificially corrupted films, and the experimental results showed that the proposed algorithm is very efficient and the quality of the images reconstructed using the proposed method is competitive with that restored using existing film restoration methods.