Hybrid computational architectures for image segmentation

The article describes a generalizable method for creating hybrid computational architectures. This method, based on a metaphor of biological symbiosis, provides a systematic approach to combining attributes of disparate algorithms. It illustrates the approach by creating a series of hybrid architectures from a single hierarchical segmentation algorithm. Typical results from these hybrids are given. The results show that for this particular application, it is possible to leverage high-level image processing tasks with low-level algorithms. The results also demonstrate how changes in the hybrid architecture can introduce nuances in the segmentation output