Cellular Automata-Based Image Sequence Denoising Algorithm for Signal Dependent Noise

This work deals with the problem of denoising sequences of multi-dimensional images that are corrupted by different types of noise. The denoising is performed through a cellular automata based filtering structure (4DCAF) that jointly considers spectral, spatial and temporal information by means of a three-dimensional neighborhood when each pixel of the sequence is processed. The novelty of the proposed method is its capacity to contemplate information concerning the type of noise by using as training data specific image sequences to tune the algorithm. The 4DCAF structures outperform selected state-of-the-art algorithms on both single band and multi-dimensional image sequences corrupted by different sources of noise.