Abstract:
We present a novel membrane computing model for a fundamental image processing task: a parallel and distributed seeded region growing algorithm for gray images. The image pixels are partitioned in rectangular sub-images, which are modeled as complex cells and evolve via inter-cell parallelism. Pixels inside a cell are modeled as sub-cellular objects and evolve via intra-cell parallelism. The presented model is synchronous, but can be further extended to an asynchronous version. Each cell can be efficiently implemented on a multi-core or many-core architecture and cells can communicate their boundary data via messages. With a proper granularity, the whole system can be efficiently mapped to a distributed Actor system, possibly a cloud-based Actor system. This study suggests straightforward ways to model the structured grid dwarf or others of the 13 parallel algorithmic patterns collectively known as the 13 Berkeley dwarfs.