| 2. |
- Lindberg, Siv M, et al.
(författare)
-
Making subjective assessments objective : A mottle ruler for calibration of panel assessments of perceived print mottle
- 2008
-
Ingår i: Paper Conference and Trade Show, PaperCon '08. - 9781605605104 ; , s. 2086-2124
-
Konferensbidrag (refereegranskat)abstract
- The objective was to establish principles and procedures for: a) digital simulation and output of reference samples for perceived print mottle assessment, b) calibration of subjective assessments to a common unit of measurement. We decided to calibrate our scales using the just-noticeable difference (jnd) as the unit of subjective measurement and adapt a ruler system for the assessment of perceived print mottle as suggested by Keelan [1]. Jnd used in this context means the smallest difference in mottle required for 50 % of a test panel to genuinely detect a difference. Reference samples were designed in Matlab by adding a stochastic noise to a homogeneous gray (33 % reflectance) and a cyan image, that, when printed would have the resemblance of print mottle typically seen in offset printing. Digital images were converted to CMYK images, rasterized using a Best Color Rip, and printed on Epson Archival Matte paper on an Epson Stylus Photo 2100 ink-jet printer. The resulting prints were evaluated for the magnitudes of perceived visual mottle as well as for the just-noticeable differences, jnds, using panels of assessors. The prints were also evaluated for instrumental mottle using STFI-Packforsk’s image analysis based STFI-mottle software. Two mottle rulers were assembled. A mottle ruler is a set of 5 cm ×—5 cm reference samples, varying widely in magnitudes of reflectance variation resembling real print mottle. Rulers were constructed for simulated K40 % (gray) and cyan 100 % (cyan) samples. The rulers are arranged with approximate twelve reference samples separated by intervals of three jnds each. A Round Robin evaluation of visually perceived print mottle was performed. Two rulers, cyan and gray, together with test samples to be evaluated against the rulers, were sent out to four participating companies. The test sets consisted of eight commercially produced samples (printed paper and board) and twelve simulated samples varying not only in the magnitude of mottle but also in character (e.g. coarseness and the orientation of the simulated noise). The purpose of using test samples with appearances that differed from the reference samples was to test the performance of the approach. The mottle rulers were also applied at Iggesund Paperboard during a test period of eight months in which more than 300 printed samples were assessed against the rulers. Round robin results showed good agreement between participating paper companies. The average standard deviations of the ratings against the rulers were ˜2.5 jnd units for the gray samples and ˜3.5 jnd units for the cyan samples. Some difficulties with the assessment technique were observed for the simulated samples, due to some samples having a very different visual character in terms of uncovered areas or noise that had a distinct orientation e.g. banding. Results from the testing period at Iggesund show less variability; average standard deviation for cyan 1.61 jnd and for gray 2.3 jnd. Overall, the calibration approach developed works very well, with the results showing a low variability of the ratings and a very good agreement between the different test panels, as well as a reasonably good correspondence between ratings of commercially produced samples and mottle measurements.
|
|
