A Three-valued Logic Approach for Edge Detection Using Cellular Automata
Manuel P. Cuéllar, Ramón Rueda, Luis G. Baca Ruiz and María Del Carmen Pegalajar
In this paper, we propose a novel method of edge detection in grayscale images using cellular automata. We use genetic programming to train Straight Line Programs that model Well-Formed Formula in a three-valued logic system. These formula are the rules that govern the behaviour of cellular automata. Unlike previous approaches of cellular automata for edge detection, we take advantage of explicit gradient direction information to find the best rule that accomplishes the task. Comparisons with previous works in the experimentation are carried out using the USF benchmark dataset, and show that the proposal overcomes some limitations of state-of-the-art cellular automata methods for edge detection.
Keywords: Cellular automata, edge detection, genetic programming, straight-line programs, 3VL