Printer calibration with improved stability and inter-printer consistency
Publication Date: 2005-Oct-03
The IP.com Prior Art Database
Normal printer calibrations linearize (in some space) the printer response over the entire range from white paper to full density. If the full density value changes from time to time, or from printer to printer, the end point of the line changes and the printed output varies accordingly. This reduces temporal (e.g., day to day) consistency and print-engine to print-engine consistency. This dosclosure describes solutions to the color stability problem.
Normal printer calibrations linearize (in some space) the printer response over the entire range from white paper to full density. If the full density value changes from time to time, or from printer to printer, the end point of the line changes and the printed output varies accordingly. This reduces temporal consistency and print-engine to print-engine consistency.
A maximum density value is chosen to be achievable across an extended period of time and across multiple printers (of the same model). This value might be directly achievable 90% of the time. Measured values are then adjusted to limit the range of each printer at each calibration to only use values that are within the range of 0 to the maximum density. If the maximum density measured for a given calibration is less even that the chosen maximum, extrapolation is used to obtain the end value.
Printers are generally calibrated to:
• bring them to a known state
• bring them to a state in which the device color space is well behaved, so that inverting the characterization is relatively easy
Typically the latter is emphasized over the former. Two common methods of calibration are known as DE from paper and gray balance tone reproduction curves or TRC. In US Patent publication 20040257595, and “Two-dimensional Transforms for Device Color Calibration” [proc. IS&TISPIE,
San Jose, 2004], Sharma et al describe a method that combines the two. That method would be improved with the method described here, as it is an improvement on either one.
DE from paper
Patches are printed along each of the four lines running from white paper to full on of each of the four primary colors. For each patch, the Euclidean distance in L*a*b* space from paper is computed. For each primary color independently, a curve is computed mapping from 0 coverage of the primary, and 0 distance from paper to maximum coverage and maximum distance from paper. The distance from paper is linearly scaled so that the maximum distance from paper maps to 255 (if the scale being used is 0 to 255). The curve is inverted, and a new curve is found that matches the inverse of the given curve and maps 0 to 0, and 255 to 255. Thus the resulting TRC will always have output value ranging from 0 to 255, and the spacing of those values will make the steps as even as possible in units of L*a*b* distance.
Grey Balance TAG
An initial model of the printer behavior is used to generate CMY only patches close to the neutral axis or any preferred aim curve mapping L* to (a*,b*) pairs. These patches are measured, and their errors from the aim are calculated. These combined with the initial model give a better estimate of which CMY values to use to achieve the aim curve. If the errors (and hence refinement) are small, this refined estimate is used directly. Otherwise it is used to produce a new set of patches which are printed and measur...