Functional-level development of image filters by means of cellular automata

A developmental method based on one-dimensional uniform cellular automaton is presented for generating image filters at the level of functional blocks. The key idea is to enhance the local transition function of cellular automaton in order to enable its cells to generate functional blocks when determining the new states during development. Simple genetic algorithm is applied to find a suitable cellular automaton (its initial state and the transition function) that is able in a finite number of steps to generate a functional structure for image filtering. Several sets of experiments are presented considering various settings of parameters of the developmental system. The evolved filters are evaluated using different types of grayscale images corrupted by salt-and-pepper noise of various intensity. The obtained filters are compared to some conventional median filters with respect to the filtering quality.