Application of Totalistic Cellular Automata for Noise Filtering in Image Processing

摘要

The selection of the neighbourhood is a very important part of the specification and training of Cellular Automata (CA) in image processing. Rather than guessing or assuming a specific neighbourhood, this paper investigates the selection of the neighbourhood and studies how the level of added noise in the image affects the selection of an optimal neighbourhood. To enhance the performance of noise removal using Cellular Automata, a basic totalistic CA (BTCA) model and a new weighted totalistic CA (WTCA) model are introduced. Both methods require much less memory storage and are feasible in practice even for very large neighbourhoods. Several experiments are presented to demonstrate that both proposed methods produce consistently better performance than the median filter and the traditional CA method for low noise levels, and for filtering at high noise levels, the WTCA model is shown an excellent performance compared to other methods.