Digitization and Metric Conversion for Image Quality Test Targets

A common need of the INCITS W1.1 Macro Uniformity, Color Rendition and Micro Uniformity ad hoc teams is to digitize image quality test targets and derive parameters that correlate with image quality assessments. The digitized data should be in a colorimetric color space such as CIELAB and have no spatial artifacts that reduce image quality parameter accuracy. Input digitizers come in many forms including inexpensive scanners used in the home, a range of sophisticated scanners used for graphic arts and scanners used for scientific and industrial measurements (e.g., microdensitometers). Some of these are capable of digitizing hard copy output for image quality objective metrics, and this report focuses on assessment of high quality flatbed scanners for that role. Digitization using flatbed scanners is desired because they are relatively inexpensive, easy to use, and most are available with ADF permitting analysis of a stack of documents with little user interaction. Other authors have addressed using scanners for image quality measurements. This paper focuses on color transformations from RGB to CIELAB and demonstrates that flatbed scanners can have a high level of accuracy for generating accurate, stable images in the CIELAB metric. Flatbed Scanners as Input Digitizers for Image Quality Work The scanners considered herein have been designed for the graphic arts industry. Their accuracy in the role of image quality digitizers is dependent upon a number of factors, including: 1. Scanner spatial uniformity. Two types of uniformity were assessed: a. Overall top-bottom, side-to-side uniformity. This can be assessed with uniform input targets, such as a sheet of Munsell neutral. For a good quality scanner, variation within a letter-size page should be a fraction of one RGB code value. Very highresolution scans that increase scan times significantly, can reduce uniformity because of lamp and electronics changes. b. Pattern noise from sensor and/or motion noise. These was assessed both visually and using Fourier analysis to detect patterns within the image. For good quality scanners spatial uniformity was not an issue. 2. Scanner repeatability. Including samples of Munsell material along the edge of each scan can monitor scanto-scan variability. Warming up the scanner before each image capture can reduce variability and typically, scan-to-scan differences have been a fraction of one RGB code value. 3. Scanner resolution. Low scan resolution can cause aliasing with output device halftones. For microuniformity work, scanner internal down-sampling methodology may generate problematic data because of data elimination instead of averaging. For critical micro evaluations, highest optical-resolution scans and averaging down in linear space may produce the best results. For this study, most materials were continuous tone and all reported results were from scans done at 300 dpi. Very similar results were obtained for 600 and 1200 dpi scans. Scanner MTF was characterized using Sine Patterns test targets and was found to be more than sufficient. 4. Automatic exposure adjustment. Some scanners automatically adjust factors such as exposure, color balance and contrast to reduce user effort in obtaining high quality images. This can both increase variability and eliminate signal (e.g., cause saturation) in image quality measurements. Automatic color settings and use of color conversions should be turned off or disabled and previews watched for automatic changes in density and balance. The use of uniform samples of Munsell materials such as RGBCMY primaries and neutrals can monitor any automatic adjustments. These can be placed (e.g., taped) outside the normal scan area and included in all scan data. 5. Scanner dynamic range. Lower cost scanners may lack the dynamic range required for accurate digitization of test targets. Relating target progressions in neutrals and colors to measurements with GretagMacbeth or X-Rite spectrophotometers can test dynamic range. 6. Uniformity of the hard copy output being characterized. Poor uniformity from irregular media (e.g., textured), the marking process (e.g., ghosting, holes, adjacency, halftoning, granularity) and/or dust/dirt on the print can introduce variability in both colorimetric and scanner measurements. 7. Integrating Cavity Effects, or ICE. Flatbed scanners can have considerable adjacency effects introduced by target reflections back into the illumination system and Procedure for this Analysis then directed back onto the document. The reflectance measured by the sensors is not only a function of the document reflectance at that point but of surrounding reflectance. Scanner design, target design and scanner processing can influence the degree of ICE effects in a scan. ICE is discussed in the next section. An image quality assessment includes a printing of all image quality test targets at the same time, the same printer settings and using the same media. The targets are scanned, RGB data is converted to CIELAB and image quality metrics are calculated. Figure 1 shows a flow chart of the characterization and analysis procedures used for this study. 8. Bit-depth. It is important to retain accuracy throughout the image quality assessment procedure and 16-bit capture is recommended. All analysis performed herein has been done with 16-bit RGB capture converted to 16-bit CIELAB values. 9. Color accuracy. Calculations of metrics correlated with appearance require that RGB scanner data be converted to a color space such as CIELAB. Because scanners are not colorimetric, accurate conversion of RGB scans to the CIELAB metric requires a target made from the particular hard copy device (colorants and media) being assessed. The conversion should not generate spatial artifacts. This paper demonstrates that flatbed scanners can satisfy all these conditions and are capable image quality digitizers. Integrating Cavity Effects, ICE The adjacency effects of ICE were large enough in the initial experiments using inexpensive flatbed scanners to warrant additional investigation. Experiments with wedge (“V”) targets and different reference density levels, and black masks over parts of the document demonstrated that (1) the ICE influence can extend a number of millimeters, (2) the influence may not be symmetrical in the horizontal and vertical scan direction, (3) high density strips between patches appear to reduce these effects, (4) the effects are different than flare, an overall veiling glare affecting mainly darker areas of the document, and (5) if the scanner has ICE compensation, it may not be perfect. The conclusion was that high quality scanners and neutral separations between the patches can reduce, but not eliminate, ICE. image f IT8.7/2 Target (photo media)