多元回归法是彩色打印机特性化普遍采用的方法,实验对影响多元回归精度的阶数、色空间选择、以及训练样本选取进行了研究。研究结果表明:(1)二、三、四阶的特性化精度分别在10.9±0.3、6.6±0.2、6.0±0.2个CIELAB色差单位;(2)回归模型采用RGB到CIELAB或CIEXYZ的预测精度基本相同,大概相差1个CIELAB色差单位,但如果训练样本选择适当,两种色空间无差距;(3)训练样本的RGB取值要包括打印色域边界点0和255,这样将保证模型总体预测精度。研究结果将指导实际应用中回归阶数、色空间选择和训练样本的选择。
Polynomial regression is a current method of characterizing color printers.The order number,color spaces and training samples of the polynomial regression are studied further on a color inkjet printer.Experimental results show:(1) The characterizing precision of the polynomial regression with second-order,third-order and fourth-order are respectively 10.9±0.3,6.6±0.2 and 6.0±0.2ΔE*ab;(2)The difference of characterizing precision of the polynomial regression by using RGB to CIEXYZ or CIELAB is about one ΔE*ab,and with increase of the training samples,this difference disappear;(3)The R,G,B of the printer should include 0 and 255 in the training samples,because both are boundary points of color gamut of a color printer,which will improve total characterizing precision of the polynomial regression model.