3种主流分色算法的误差分析

Errors Analysis of Three Major Color Separation Algorithms

在分析多项式回归、神经网络、修正的纽介堡方程这3种主流分色算法基本原理的基础上,对这3种类型算法分别采用新的实现方法进行设计,即:多项式回归算法的建模,通过多元回归分析求解出多项式的最小二乘估计值进行;神经网络算法的建模,采用基于径向基的RBF神经网络;纽介堡方程的建模,采用迭代法直接求解的方式。最后采用Matlab编程实现,并实验比较了3种算法的分色精度、稳定性等性能。研究结果表明:多项式回归算法稳定性和精度都最佳且平均色差小,RBF算法色块建模速度快,转换效果较好。最后分析了分色结果,提出了算法改进方向。

The fundamental of three major color separation algorithms was analyzed,of which were Neugebauer equations,polynomials,and neural networks.The algorithms were designed with new methods： polynomials were solved by multiple regression analysis;neural networks were modeled by RBF;Neugebauer equation by iteration method.They were programmed by Matlab and experiments were designed to compare the accuracy and stability of the three algorithms.The results showed that polynomial regression algorithm＇s stability and accuracy were the best and has the smallest average color difference;RBF algorithm was the fastest.Improved method was proposed by analyzing separation results.