摘要
目的真网点打样使用调频网点来模拟印刷调幅网点,能在数码打样上获得与印刷调幅加网相同的参考效果。方法利用BP神经网络获得数码打样与印刷输出的CMYK-CMYK网点面积率的非线性映射来实现真网点喷墨打样。结果基于BP神经网络的真网点喷墨打样,识别率稳定在95%左右,训练100次左右就能达到收敛。真网点形态变形小,色域映射准确,与标准胶印印张的色差均值为2.24,色差标准差为1.47,色差方差为2.16,色差熵为3.12。结论基于BP神经网络的真网点数码打样和标准胶印印张的网点面积率转换不需要精确的数学模型,具有原理简单、转换迅速和适应性强等优点。无论在网点外观、色域和色差各方面,都比传统数码打样具有更多的优势。
The work aims to provide reference for promotion of AM screening technologies on digital proofing based on the true dot proofing that uses FM dot to imitate printing AM dot. BP neural network was used to obtain the nonlinear mapping of digital proofing and CMYK-CMYK dot area coverage outputted by the printing to achieve true dot inkjet proofing. The recognition rate of the true dot inkjet proofing based on BP neural network was stabilized at 95% or so, and the training for about 100 times made it possible for convergence. The true dot deformation was small, the color gamut mapping was correct, the average value of color difference with standard offset printing sheet was 2.24, the standard deviation and the variance of color difference were 1.47 and 2.16 respectively, and the color difference entropy was 3.12. The conversion of the true dot digital proofing based on BP neural network and the dot area coverage of standard offset printing sheet requires no accurate mathematical model and it is characterized by such advantages as simple principle, rapid conversion and strong adaptability. Compared to the traditional digital proofing, the true dot digital inkjet proofing has more advantages whether in dot appearance, color gamut or color difference.
出处
《包装工程》
CAS
北大核心
2017年第3期175-179,共5页
Packaging Engineering
关键词
数码打样
真网点打样
BP神经网络
色差
digital proofing
true dot proofing
BP neural network
color difference
