Abstract

We have recently proposed and demonstrated optical neural networks (ONNs) with liquid crystal televisions (LCTVs), in which usually color LCTVs are used.^1,2 Since each color pixel is composed of red, green, and blue elements, the resolution is reduced. However, the color LCTVs offer the capability of storing and processing color information. A polychromatic ONN is presented in this paper. In the training phase, the color images are decomposed into three primary colors. The monochrome interconnection weight matrices (IWMs) are constructed based on three independent training sets in primary colors. The polychromatic IWM is then composed and displayed on the LCTV for the reconstruction of the stored color patterns.

© 1991 Optical Society of America