Modular neural network-type CANFIS neuro-fuzzy modeling for multi-illumination color device characterization

摘要

This paper describes adaptive-network modeling for color correction/compensation through multi-illuminant color device characterization of an electronic video camera. In particular, we emphasize a great potential for practical use of modular neural network-type CANFIS neuro-fuzzy models and their advantage over a single MLP approach as well as conventional lookup-table-based (ThC-matrix) methods by demonstrating their remarkable approximation and generalization capacity even when they are optimized with only four-illuminant data.