Methods
Varnish Mixture
The Regalrez and PVA were chosen for the difference in molecular weight between the two. The PVA has a much higher molecular weight than the Regalrez. The Regalrez was mixed 1 to 1 ratio with paint thinner by weight. The PVA was mixed 1 part PVA to three parts Arcosolv PM solution. These mixtures attempt to take the difference in molecular weight into account so that when each layer dries, the layers will be of an equal thickness.
Target Creation
It was necessary to create test targets to which the varnish could be applied because a major idea behind this experiment is that the data be taken from reflection objects. For the spatail analysis, slant edge targets were made by applying a layer of matte white to a smooth card and then an edge of approximatley five degrees from normal was made with black paint. The target for the colorimetric analysis was made from Sherwin Williams latex house paint, chosen for its dark colors and matte properties, applied to a regular canvas. Varnish was applied to the targets using a .003 inch draw-down bar. These methods were chosen to help simulate actual working conditions.
Image Acquisition
In order to image the targets, an IBM Pro – 3000 camera was used. This camera is a line scan camera that generates a 12-bit image per color channel.
Colorimetric Evaluation
The first step in the colormetric evaluation was to determine the actual X, Y, and Z values of the Gretag Macbeth ColorChecker Color Rendition Chart being used. This was done by integrating the spectral reflectance of each patch with the spectral output of the lamps illuminating the target and the CIE color matching functions for the 1931 two degrees standard observer and then multiplying by a normalization constant. This integration was performed over an interval of 380nm to 730nm at 10 nm steps. The following equations were used:
= 10nm
where Sl is the spectral output of the source illumination, Rl is the spectral reflectance, and x,y,and z bar are the appropriate CIE color matching function.
The next step was to determine the values which would linearize the digital counts of the camera. This was done by applying the following gain–gamma-offset equation to the digital counts given by the camera and then minimizing the RMS tristimulus error of the linearized digital counts of the Macbeth grayscale.
where R=raw digital count, B=max digital count (2^bit depth –1), G=gain.
These values were then used to generate a 3x9 pseudo inverse matrix that could be used to transform linear digital counts to CIE XYZ space.
Third, the XYZ values of the unvarnished and varnished color target were found. These values were then converted to CIE L*a*b* space and the color difference can be found using the following equations:
where the unvarnished target was the standard and the varnished target was used as the measured.
Spatial Evaluation
In order to evaluate the SFR of the slant edge targets with the Image Analyzer plug in, the 12 bit camera data had to be converted to 8 bit and a new lookup table for the plug in had to be created. This was necessary because the plug in was designed only to accept 8 bit input. After this was done, the image’s SFRs were found in the green (luminance) channel. The CMT Acutance of the varnished and unvarnished target for each specific resin was then computed.
Then,
the human Contrast Sensitivity Function was generated using the following
equation proposed by Mannos and Sakrison[6] for its ease of analysis.
f = frequency in cycles per degree
In order to assign a quantitative number to the varnish, these curves are converted to cycles per millimeter, multiplied point by point and integrated. This yields the CMT Acutance.
For the CMT Acutance, a range of 0 to 9 cycles per millimeter was used. Anything outside this was considered noise.