Self-reconfigurable evolvable hardware system for adaptive image processing

摘要

Inspired by concepts from biology, applications for customizable systolic architecture for image processing based on field-programmable gate arrays (FPGAs) often use unintentionally deceptive terms to convey some modestly significant results. Evolution (in theory) occurs over almost geological time scales, which is not the time scale of the hardware systems for these projects, nor of their progressively converging software heuristics, nor of their likely neural network-inclusive future generations. Despite this, evolution really seems to be the dominant metaphor for conceptualization and expression in this field. Peeling away these nomenclature wrappings, the authors of the paper direct "adaptive image processing" to a trade-off between noise filtering and edge detection, not to recognition, nor to understanding.