A cellular automata-based filtering approach to multi-temporal image denoising

Abstract

This work tackles the challenge of denoising image sequences by employing a spatio-temporal filtering approach based on cellular automata. The proposed algorithm, known as st-CAF, is distinguished by its use of a spatio-temporal neighborhood when processing each pixel in the sequence. Another significant feature of this method is the evolutionary process used to determine the rule sets for the cellular automata, enabling effective adaptation to various types of images and noise through the use of a suitable training set. This adaptability provides a significant advantage over traditional single-frame denoising methods described in the literature, as well as their adaptations for sequences. The paper highlights this advantage by demonstrating the algorithm's performance on different types of noisy images and comparing it with other techniques.