École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland.
IEEE Trans Image Process. 2011 Feb;20(2):513-22. doi: 10.1109/TIP.2010.2063037. Epub 2010 Aug 9.
Today's spectral reflection prediction models are able to predict the reflection spectra of printed color images with an accuracy as high as the reproduction variability allows. However, to calibrate such models, special uniform calibration patches need to be printed. These calibration patches use space and have to be removed from the final product. The present contribution shows how to deduce the ink spreading behavior of the color halftones from spectral reflectances acquired within printed color images. Image tiles of a color as uniform as possible are selected within the printed images. The ink spreading behavior is fitted by relying on the spectral reflectances of the selected image tiles. A relevance metric specifies the impact of each ink spreading curve on the selected image tiles. These relevance metrics are used to constrain the corresponding ink spreading curves. Experiments performed on an inkjet printer demonstrate that the new constraint-based calibration of the spectral reflection prediction model performs well when predicting color halftones significantly different from the selected image tiles. For some prints, the proposed image based model calibration is more accurate than a classical calibration.
如今的光谱反射预测模型能够以与再现可变性允许的精度相当高的方式预测印刷彩色图像的反射光谱。然而，为了校准这些模型，需要专门打印特殊的均匀校准补丁。这些校准补丁需要占用空间，并且必须从最终产品中移除。本贡献展示了如何从印刷彩色图像内获取的光谱反射率推导出彩色半色调的油墨扩展行为。在印刷图像内选择尽可能均匀的颜色图像块。通过依赖于所选图像块的光谱反射率来拟合油墨扩展行为。相关性度量指定了每个油墨扩展曲线对所选图像块的影响。这些相关性度量用于约束相应的油墨扩展曲线。在喷墨打印机上进行的实验表明，当预测与所选图像块明显不同的彩色半色调时，基于新约束的光谱反射预测模型的校准效果良好。对于某些打印件，所提出的基于图像的模型校准比经典校准更准确。
