JAIC 2002, Volume 41, Number 1, Article 4 (pp. 43 to 60)
JAIC online
Journal of the American Institute for Conservation
JAIC 2002, Volume 41, Number 1, Article 4 (pp. 43 to 60)




The method for measuring paint swelling proposed here is a relatively low sensitivity technique suitable for use on unsupported paint films. Lateral, in-plane swelling is determined by a simple method that is loosely based on that of Brunt (1964) and Eissler and Princen (1968, 1970) but with significant improvement to the measurement aspect through the use of computer-based image analysis.

The general experimental setup is illustrated in figure 1. Images of the paint samples during immersion in solvent are captured using a video camera fitted without lenses, via a trinocular viewing head, to a low-power stereo zoom microscope (objective range 1–4x). The video camera is linked directly to a computer: images are captured and digitized by an internal image grabber (Neotech). Two video cameras were used for the results reported here: a Moritex Micro-Scopeman MS-500HR-RGB and a JVC TK-C1380E. Both have 752 x 582 pixel array detectors. The computer used was an Apple Power Macintosh G3 computer (233 MHz, 4 GB disk, 128 MB RAM, 6 MB SGRAM, 20 in. Sony Trinitron monitor). The digitized images are in monochrome, 8-bit gray scale, i.e., a maximum palette of 256 levels of gray. All raw and processed images are archived onto a rewritable disc.

Fig. 1. Experimental setup for conducting the swelling experiments

The paint fragments for testing are cut from larger samples of known thickness and are placed on a disc of glass-fiber filter paper. In turn, this disc is placed in a semi-opaque glass basin in which the swelling test is conducted. The basin is viewed through the microscope in transmitted light using the microscope's internal tungsten-halogen light source, the semi-opaque basin and filter paper serving as diffusing screens so that the paint fragments are seen as silhouettes against a uniform, bright background. To limit heat gain from the tungsten-halogen light source, the lamp is switched on only for the few seconds it takes to capture an image. The paint samples are covered with a glass slide supported on glass spacers 0.5–0.6 mm thick, sufficient to give the samples clearance even in their swollen condition. This arrangement effectively creates a cell of constant dimension that is filled with solvent during the swelling experiment. In some of the early experiments reported here, the glass cover slide was a conventional stage micrometer specifically designed for viewing in transmitted light. Later this slide was replaced by a custom-designed slide chromium-printed with a calibration scale and two semicircular reference shapes, the latter feature thus precluding the need to include a separate reference particle in the experiment cell. The cover slide serves as the reference measure against which the images are calibrated. The inert reference particles are included in the captured images as internal standards in order to detect any temporal variation in image constancy and to provide a means of normalizing the paint fragment areas should temporal variation occur to a significant degree. Either the custom-designed slide mentioned previously is used for this purpose, or copper discs, ca. 3 mm in diameter, cut into segments.

To conduct a swelling experiment, the selected solvent is quickly run into the basin from a dropping funnel. Care must be taken not to displace the paint fragments from the field of view; this step can be difficult with solvents having high surface tension, such as aqueous solutions. The first image, representing t = 0, is captured the moment the paint fragments become wetted with solvent in the meniscus between the basin and the stage micrometer. The t = 0 image defines the initial paint fragment dimensions from which changes in area are determined. As they swell in the solvent, images of the samples are then taken at regular intervals over a time period ranging from 40 to 60 minutes for the faster-diffusing solvents like acetone, to 180 minutes or more for slower diffusers, or even for periods as long as 24 hours for aqueous liquids. Focus and magnification are not adjusted during an experiment in order to ensure internal consistency of images, and the scale of the cover slide is included in the field of the images for calibration purposes. All images of the samples immersed in solvent are captured ostensibly under constant conditions of illumination and viewing. As mentioned previously, the microscope lamp is illuminated only when necessary to capture an image of the samples in order to prevent heating the solvent and paint samples. During the course of the swelling experiment, the basin is covered to restrict evaporation and evaporative cooling of the samples. The initial volume of solvent in which the samples are immersed is generally on the order of 10 ml. No additional solvent is added to the basin during the course of the swelling experiment for reasons that are explained below. The stock of test solvent is always conditioned to the ambient room temperature to limit any thermal effects.

Clearly, the absolute resolution of area measurement will be dependent on the degree of magnification employed for a given experiment. There are, however, advantages to using lower magnifications, for lower magnification provides a broader field of view through the microscope and, consequently, the possibility of including more sample fragments in a single swelling experiment. The choice of magnification is, therefore, something of a compromise of dimensional resolution and quantity of data/statistical validity. In practice, differential sample response is by far the strongest cause of variability in measured results. For the majority of experiments conducted so far, magnifications of 1.5–2.0x are generally employed, which allow 8–10 paint fragments ca. 2–5 mm2 to be measured for swelling simultaneously and with reasonable accuracy. In terms of absolute resolution, at an objective magnification of 2.0x, one pixel corresponds to an area on the order of 0.000124 mm2. Higher magnifications are, of course, perfectly possible, but these give reduced fields of view and so constrain the size or number of samples that may be viewed simultaneously. Where possible, following the approach of Eissler and Princen (1968, 1970), the small paint fragments under test are triangular in shape. However, the more brittle films often fracture into shapes other than triangular, but paint fragment shape has been found to exert little influence on the area-swelling measurements. All paint samples of a given type are substantially the same thickness. Thickness ranges of the paint films examined are given in table 1.

The raw data, then, from a swelling experiment are a set of images of paint fragments at different times and degrees of swelling. A typical raw image is shown in figure 2a. For strong-swelling solvents, the dimensional changes that samples undergo during the course of swelling are clearly perceptible on a purely visual basis. Typically, 10–15 images are collected per experiment, with images captured more frequently in the early stages when the rate of swelling is greatest. All the sample images thus captured are image-processed as a group: the same operations are applied to all images in an experiment set. The first stage in the electronic processing is adjustment of brightness and contrast to render the paint fragments as distinct dark particles against a white background, which is done in Adobe Photoshop 5.0 (see fig. 2b). The same brightness and contrast levels are applied to all the images of immersed samples. Sharpening filters may also be applied to the digitized images if thought to improve image quality. Again, the same filter is applied to all images in a set to ensure internal consistency.

Fig. 2. Image-processing operations: a, immersed samples, raw image; b, brightness, contrast levels adjusted; c, calibrated and thresholded. Area measurements in Optilab 2.6.1 are made on the calibrated and thresholded images.

Quantitative data on the size of the individual paint fragments cannot be extracted directly using Adobe Photoshop, and to measure the changes in paint fragment dimensions the images are exported to Graftek Optilab 2.6.1. This software package has been used previously to measure changes in weave geometry parameters of artist's canvases as a function of RH (Bilson 1996). The images of paint fragments immersed in solvent are first calibrated for dimension, as a group or individually, against the stage micrometer. They are then thresholded to render the paint fragments as discrete particles against a dark background, as in figure 2c. The Optilab software has functions for detecting discrete particles and for defining their dimensions in terms of more than 20 different parameters. At the resolutions operating in the present method, sample perimeter as an indicator of swelling—which was determined microscopically in the method of Eissler and Princen (1968, 1970)— was found to involve a high degree of variability due to differential pixelation in the image capture and processing operations. Particle area has been found to be a far more reliable descriptor of paint fragment swelling. The absolute values of particle area measured at each time interval are converted to % change in area by simple calculation against the t = 0 values

Table . Summary of Composition and Condition of the Various Test Paint Films
using the equation:
The data are presented as plots of time vs. % change in area (ΔA%). Results for an experiment involving several paint fragments of the same type are combined to produce a single mean swelling curve.

Copyright � 2002 American Institution for Conservation of Historic & Artistic Works