Print Quality Inspection
Digital presses print high quality material, as such quality assurance is imperative. The faster upstream a print defect is detected, the cheaper it will be to correct the problem. Many print shops have multiple quality control stations some of them will have people look at pages coming off the press. Today’s digital presses are fast and print double sided. This alone makes direct quality control problematic. In addition, the workflow is automatic which means that multiple jobs will print in sequence and be handled in a streamlined manner, which complicates the situation even further.
HP Labs Israel developed the first Quality Assurance system for digital commercial printing. Indeed automated inspection for fast offset presses is a known technology, but it relies on golden templates which will print once, be approved, and then be used as a reference to multiple identical consecutive prints. This practice will not work for digital printing where runs of identical prints are very short, and often every page is different.
The potential for online true digital quality assurance as outlined by HPL-Israel researchers motivated Indigo’s R&D to include a scanner in the paper path of future presses. Indeed today’s HP-Indigo presses have now an inline inspection option, and the scanner is used for many other differentiating tasks such as automated calibration and diagnostics. The latter enables automated faster resolution of print quality issues by non-expert press operators.
|The End Game|
Technology developed at the lab for automated print quality diagnostics was sold as part of a product with the introduction of the inline scanner in May 2010. The digital quality assurance technology was introduced as a major new feaure of the HP Indigo 7600 press in May 2012.
The Inspection technology and system developed
at HPL Israel drives Indigo’s Automated Alert Agent
The inspection technology behind the Automated Alert Agent is all about perceptual image difference. We have to compare the image of the digital original and the scanned image of the actual print. But, images should be compared as a human would, namely, we have to understand what differences are expected due to the print and scan process, what differences are ‘forgivable’, and modify this decision accordingly. Notice that some of the decisions depend on the job print content and some on customer requirements. To that end visual feedback on customer jobs needs to filter ‘technical’ differences, and be able to scale all other differences perceptually.
The image of the AAA interface depicts a defect detected and reported by the system
The interface of the AAA system reporting a defect
The system behind AAA is not trivial nor is it simple. It includes known and novel algorithms spanning color processing, image registration, mathematical morphology, and parallel computing. However, the technology which made the difference is our image difference technology. It is based on the state of the art Structural SIMilarity (SSIM) metric, but makes sure it is invariant to sub-pixel misregistration inherent in the print and scan scenario.
Understanding that diagnostics is changing with the machine it diagnoses, we developed a technology that can learn to identify new defects based on tagged examples so that new diagnostics routines can be automated based on a simple procedure rather than involving an imaging expert every time there is a need to update print quality troubleshooting. The print quality diagnostics technology is based on perceptual quality evaluation and machine learning.
Learning Print Artifact Detectors
Hila Nachlieli, Hadas Kogan, Morad Awad, Doron Shaked, Smadar Shiffman (2012). 6th European Conference on Colour in Graphics, Imaging, and Vision. IS&T Amsterdam, May 6-9, 2012
GPGPU@ HP Research
Automatic visual inspection and defect detection on Variable Data Prints
Vans, M.; Schein, S.; Staelin, C.; Kisilev, P.; Simske, S.; Dagan, R.; Harush, S. (2011). Journal of Electronic Imaging (January 1, 2011), 20(1), 13010-13010-13. DOI: 10.1117/1.3537837
Perception Guided Automatic Press Diagnosis
Nachlieli, H.; Karni, Z.; Raz, S. (2011). NIP27: 27th International Conference on Digital Printing Technologies and Digital Fabrication, October 2-6, 2011, Minneapolis, Minnesota, USA, 784-787.
Measuring the quality of quality measures
Nachlieli, H.; Shaked, D. (2011). IEEE Transactions on Image Processing (January 1, 2011), 20(1), 76-87. DOI: 10.1109/TIP.2010.2059708
A robust similarity measure for automatic inspection
Automatic Mechanical-Band Perceptual Evaluation