Cellular automata for image noise filtering and edge detection
Олександр Жежерун, Bohdan Kalitovskyi · NaUKMA Research Papers Computer Science · 2019
Cellular Automata (CA) are the most common and simple models of parallel computations. CA can be successfully applied in image processing, where we consider images as a system of simple components (pixels), and the behaviour of each component is obtained and reformed according to the behaviour of their neighbours and their previous behaviour. The constructive components of these systems can perform reliable and complex tasks by interacting with each other. Precisely by setting certain rules of the behaviour of the components, the cellular automata achieved significant results in such areas of image processing as noise filtering, smoothing, edge detection, restoring and extracting the features of images, figures and texts recognition, image compression. However, up to these days corresponding researches remain being used only in order to solve specified tasks, such as image processing of minefields, the processing of X-ray images in medicine, or the analysis of satellite imagery. This paper reviews the application of CA for image analysis and processing. It demonstrates an image noise filter based on CA, which can remove impulse noise from a noise-corrupted image and compares it with the median filter. Meanwhile, the edge detection appears to be one of the most crucial tasks in image processing (especially for biological and medical images processing). So CA based edge detection has potential benefits over known traditional approaches since it is computationally efficient, and can be tuned for specific applications by appropriate selection or learning of rules. Several CA based edge detection methods are implemented and tested to enable an initial comparison between existing traditional methods (the Roberts cross operator, Sobel-Feldman operator, Laplace operator). This comparisons show that the provided CA-based methods are very perspective for impulse noise filtering and image edge detection.