The known approach to noise evaluation predominantly consists of an image differentiation and subsequent filtering. This approach requires an interactive procedure of pointing to homogeneous regions where the local variance of the processed image should be computed that is unacceptable in many on-line printing systems. The proposed model-based approach to noise estimation considers noise as stochastic-deterministic fluctuations of image intensity with respect to the assumed polynomial intensity function with added noise. The original image is restored from the printed image by using a robust filtering procedure based on the assumed polynomial regression model. The difference between the original image and the scanned image during the printing process is then analyzed in order to detect the significant values of the residuals which represent the printing defects or noise.
Roman M. Palenichka, "Structure-Adaptive Evaluation of Noise in Images for Automatic Image Quality Control" in Proc. IS&T Int'l Conf. on Digital Printing Technologies (NIP14), 1998, pp 586 - 589, https://doi.org/10.2352/ISSN.2169-4451.1998.14.1.art00062_2