Uniform and Non-Uniform Cellular Automata: Some Issues and Case Studies in Computer Vision

In this paper we present some low-level computer vision tasks based on uniform and non-uniform cellular automata (CAs). Uniform cellular automata are characterized by the properties of parallelism and locality, while the non-uniform CAs can perform a different local computation for every cell in the grid and they add a new degree of freedom to CA applications, preserving the most interesting properties of uniform CAs, such as synchronization and locality of connections between computing elements. We used uniform CAs to perform low-level computer vision tasks that require operations based on the SIMD paradigm (optical-flow computation and region-based segmentation) and non-uniform CAs to perform position-dependent transformations (perspective effect removal and pattern classification). A brief description of each task and its implementation oncellular automata are reported together with some discussions on the obtained results.