Assessing The Color Appearance Of Pearlescent Bottles Based On Imaging Technology

The unique color sensation of pearlescent surfaces requires complex color appearance assessment techniques, which is a fundamental approach for the color stability and high-quality color rendering of pearlescent products. Assessing the color appearance of pearlescent surface is contributive in manufacture quality evaluation, paint color matching or CAD special effect material rendering. This dissertation took pearlescent bottles as the experimental samples, aiming to establish a novel imaging method for the color appearance assessment of bottle surfaces. As a modern imaging system, the camera is competent for efficient and economical captures of color images, which can be merged to a whole robust system of online quality control for pearlescent products. The main procedures of the proposed methods included 1) multi-angle based tele-spectroradiometer (TSR) measurement and visual evaluation for the color appearance assessment of pearlescent bottles,2) digital camera based imaging methods for the color appearance evaluation, and 3) modeling of color appearance difference detecting using imaging technology.In the experiments of TSR measurement and visual assessment, the different positions were selected on the surface of the samples, which represented different viewing geometric conditions. The visual assessment using magnitude estimation of psychophysical method had acceptable repeatability, compared with the related international studies in color science. The TSR measured data were transformed to the visual attributes (lightness, colorfulness and hue), via the color appearance model of CIECAM02, and then could be compared with the data from visual experiment. The analysis results showed that they had good agreement, with the correlation coefficient R2 being above 0.90 (coefficient of variation, CV being below 20.0) for the colorfulness attribute and R2 above 0.95 (CV below 15.0) for hue; while for lightness the methods of TSR measurement and visual assessment had slope shift due to their different scales, but their linear relationship was preserved, based which, hence, a lightness linear correction algorithm was proposed in this study to reduce the CV value for lightness attribute from about 12.0-23.0 to 6.8-9.8.In order to study the color appearance assessment of pearlescent bottles based imaging technology, the polynomial model was used to characterize the digital camera for the transformation from the device-dependent RGB color space to device-independent CIEXYZ standard color space, with the mean characterization accuracy of 1.3 CIEDE2000 units on the color target of GretagMacbeth Color Checker DC. The experimental results indicated that the imaging method had better repeatability than TSR measurement because it needed less manual operations during the entire procedure. Since the bottles are curved, angle-dependent properties are already presented at some specific locations in their images captured by the camera, which could perform the rapid multi-angle measurements simultaneously. The correlation between the camera and visual results achieved R2>0.60 and CV< 24.0. The lightness correction was carried out so that the lightness had almost the same good correlation with that in hue attribute between the two methods; while the attribute of colorfulness had relatively the poorest correlation. From the data analysis of captured images it was found that the pearlescent surface had symmetrical color appearance distribution along with the symmetrical viewing geometry, and the color appearance difference between the non-glossy area and the most glossy area presented good agreement, R2=0.82 and CV=8.5, with their corresponding lightness difference.In the purpose of the color quality evaluation on pearlescent bottles, a color appearance difference detecting method was demonstrated in this dissertation. The psychophysical method of ranking order was used to determine the best two viewing geometric angles for sample-pairs exhibition, which were then applied to the visual experiment for quantifying color appearance difference through grayscale technique. Under the same conditions from the sample images captured by the camera some sub-images on the corresponding regions of different bottles were extracted, from which three sets of color parameters of glossy, non-glossy and transitional areas, as well as the lightness contrast and its transitional index via statistical analysis, were recorded as the samples'color features for both viewing geometries. Then with the lightness correction the color appearance difference detecting model was set up for pearlescent bottles. The good correlation of this model, R2=0.81 and CV=27.7, with visual inspection implied that it could be a basis for the future application in the color quality evaluation of related products with the combination of threshold setting or machine learning technique in the color computer vision system.