Image Understanding: A Driving Application for Research in Heterogeneous Parallel Processing

This paper surveys the field of image understanding, or knowledge-based computer vision, from the perspective of a driving application for research into heterogeneous parallel processing. Image understanding is so rich in its diversity of methodologies and so inherently computational intensive that it naturally demands a very heterogeneous mix of parallelism. This paper begins by exploring some of the diversity of potential parallelism in vision, and then considers the implications for both architectural and software issues of heterogeneous parallel processing.