by clicking the arrows at the side of the page, or by using the toolbar.
by clicking anywhere on the page.
by dragging the page around when zoomed in.
by clicking anywhere on the page when zoomed in.
web sites or send emails by clicking on hyperlinks.
Email this page to a friend
Search this issue
Index - jump to page or section
Archive - view past issues
FLEXO Magazine : February 2009
TECHNOLOGIES & TECHNIQUES 1 0.8 0.6 0.4 0.2 400 450 500 550 600 650 700 FIGURE 4. Spectral response of one commercially available RGB camera. duction run requirement, only relative accuracy is required. It can be assumed that during the production run, the pigments are the same, the gloss is relatively constant, the effect of scattered light remains the same (so long as you compare the same spot in the printed image on different impressions), and the UV brightener content is constant. This makes the job considerably easier since many of these sample differences that confound accuracy go away. The previous papers all focused on the absolute accuracy of RGB camera derived CIELAB measurements. I have only found a few papers that look at how accurately a camera can measure color difference. One paper reported an accuracy of color differ- ence of 2ilE; the other reported o. E. I decided to try this experiment myself. How well could a theo- retical RGB camera agree with a spectrophotometer? I started with the assumption that the fIxable sources of error would be taken care of through engineering, and that the major source of error would be the spectral response of the camera. In that way, I could start with measured spectra and compute every- thing I needed. I collected nine press sheets from a run. The sheets were run at nominal densities, with cyan low, with cyan high, with magenta low, and so on. Each sheet had 1,296 patches. The spectrum of each of these patches was measured. From the spectra, I com- puted CIELAB values and color differences between correspond- ing patches. In addition to using the spectra to compute CIELAB values, I also estimated what three commercially available RGB cameras would measure, and also estimated what a Status E densitometer would measure. Figure 4 illustrates the spectral response of one of the three RGB cameras. I used a "standard" technique to estimate CIELAB values from the RGB camera responses. I used regression to determine the optimal 3X9 matrix that would transform RGB reflectance into XYZ values. Camera derived CIELAB values were then computed for all the patches. The average color error (between CIELAB values computed di- rectly from the spectra and color values as computed through the camera response) was between l. E and 2. 3 E. This is not out of line with the previous results, but one must bear in mind that: . These are hypothetical results, assuming only spectral errors, and . This transform has been optimized for this particular stock. If this transform were to be applied to measurements with a different stock, I would expect them to be appreciably worse. But more to the point of this section, I looked at the mea- surement of color differences. Figure 5 may be helpful to understand the complicated test. The figure shows an arbi- trary patch measured on two arbitrary sheets from the seven sheets of this press run. For each patch I computed, the true CIELAB values, and also the CIELAB values that an RGB camera and appropriate transform software might report. It would be a natural test of the camera to compute the E between the actual and the camera CIELAB values. In this case, however, I computed the E between sheets as measured by the camera, and also computed the actual E between the two sheets. These two E values were then compared. FEBRUARY 2009 - www.flexography.org FLEXO
Sustainable Winter 2009