Color Information Processing based on Quaternionic Neural Networks

本研究旨在探讨可存取高维资料的复值及四元数联想记忆的分析。四元数是一类秩为四的超复数系统。主要研究了以下两个方面的内容：（1）提出并分析了一种能够存储和提取三维信号（如图像或人体坐标）的联想记忆。此外，还研究了在网络中嵌入模式的方案。(2)协同计算机是一种联想存储器，其参数采用复数值编码。通过灰度图像的检索进行性能评价。

This research subject aims to investigate the analysis for complex-valued and quaternionic-valued associative memory, which can store and retrieve higher dimensional data. Quaternion is a class of hypercomplex number system with rank four. The following two subjects are mainly explored(1) Associative memory that can store and retrieve three dimensional signals, such as images or body coordinates, is proposed and analyzed. Also, schemes for embedding patterns onto the networks are also studied.(2) Synergetic computer, which is a type of associative memory, is investigated through its parameters are encoded by complex values. The performance evaluations are conducted through the retrievals of gray-scaled images.