摘要
为了研究数码相机的颜色特征化问题,用多项式回归方法对Sony DSC F717数码相机进行了特征化分析,采用一次拟合与两次拟合分别建立RGB到XYZ空间和RGB到L*a*b*空间的多项式模型。通过色差ΔE和拟合方差σ2的比较,分析了两种模型的差异以及rgb项和常数项对模型精度的影响,并与Profile Maker软件制作的特性文件精度作了对比。结果表明:取项数为11和20时RGB→L*a*b*模型能达到较为满意的结果,两次拟合比一次拟合效果好。
To study the color characterization of a digital camera, the characterization model for Sony DSC F717 digital camera, adopting once fitting and twice fitting for deriving colorimetric mapping between the RGB signals and the CIE tri-stimulus values XYZ and L^＊a^＊b^＊ respectively, uses polynomial regression technology and is analyzed here. Through comparing the difference of AE and a2, the difference between the two models and the effect of some parameters, such as rgb and constant term on the model precision, are discussed. Compared with the ICC profile precision made by Profile Maker, the conclusion that the RGB→L^＊a^＊b^＊ model can achieve a famous result choosing term= 11 and 20, twice fitting is better than once fitting.
出处
《北京印刷学院学报》
2006年第6期9-12,共4页
Journal of Beijing Institute of Graphic Communication
关键词
数码相机
颜色特征化模型
多项式回归
digital camera； characterization modeling； polynomial regression
