Parameter nets

摘要

This paper describes the nucleus of a connectionist theory of low-level and intermediate-level vision. The theory explains segmentation in terms of massively parallel cooperative computation among intrinsic images and a set of feature networks at different levels of abstraction.