颜色空间转换的模糊神经网络辨识算法  

Color Space Transform Based on Fuzzy and Neural Network Identification Algorithm

作　　者：智川[1,2] 周世生[1] 石毅 

机构地区：[1]西安理工大学印刷包装工程学院,陕西西安710048 [2]陕西科技大学造纸工程学院,陕西西安710021 [3]陕西华拓科技有限责任公司,陕西西安710043

出　　处：《北京理工大学学报》2011年第6期722-726,共5页Transactions of Beijing Institute of Technology

基　　金：陕西省"13115"科技创新工程基金资助项目(2009ZDGC-06)

摘　　要：为了提高颜色在不同成像设备之间传递的准确性,以RGB颜色空间与CIE L*a*b*颜色转换为例,提出采用颜色空间转换模糊模型将输入颜色空间划分为多个子空间,在子空间内采用神经网络模型对输入值进行输出,利用神经网络优化颜色空间模糊转换模型,得到了基于模糊神经系统的颜色空间转换模型.研究结果表明,该模型转化精度相对于单一的颜色空间模糊转换模型有很大提高,并解决了BP神经网络颜色空间转换模型由于空间采样点数目过多而引起的训练难度增加问题.To improve the precision of color transform in different imaging devices, taking the transform of color space between RGB and CIE L ^＊ a^＊b^＊ as an example, the fuzzy model of color space transform is put forward. It divides the input color space into many subspace and, in the subspace, the input value is turned into output via neural network. The color space transform model based on fuzzy neural system is established by using neural network system to optimize the fuzzy model of color space transform. The analytical results show that the converting accuracy of the proposed model is higher than the single fuzzy model of color space transform, and the problem of training difficulty caused by much more sampling points could he solved.

关 键 词：模糊控制 神经网络 颜色空间转换 色差 

分 类 号：TP273.4[自动化与计算机技术—检测技术与自动化装置] TP301.6[自动化与计算机技术—控制科学与工程]