Abstract In this article, two Bayesian kernel methods, namely the Gaussian process regression (GPR) and relevance vector machine (RVM) techniques, are used to estimate illumination chromaticity and predict the reliability of the estimation process, which is not accessible for most machine learning techniques that have been used for color constancy. More than seven kinds of GPR covariance function and their combinations, and an RVM method using Gaussian, Laplace and Cauchy kernel functions, have been used on two real image sets. The experimental results show that the GPR method outperforms those based on RVM and ridge regression using stationary covariance functions, and GPR can almost achieve the same performance as support vector regression (SVR). The performance of the RVM for regression is almost the same as that of GPR using the dot product covariance function. The influence of outliers on the data with Gaussian noise is analyzed in detail via using heavy-tailed Laplace and Student-t kernel functions when GPR and the RVM are used for color constancy.

Abstract A series of experiments and data analyses has been performed to test the consistency between computed and perceptual color differences in images. Five International Organization for Standardization (ISO) standard color image data (SCID) images, N2, N3, N4, N5 and N7, were tested in the experiments, whose colors were altered in CIELAB lightness, chroma, and hue, either independently or simultaneously, to form the test image pairs. CIELAB, CIE94, CIEDE2000, and CMC color differences were computed by averaging the color differences pixel by pixel for the digital images or by averaging the color differences of 256 typical color patches extracted from each image for the printed images. The digital test images were displayed on an EIZO CG19 LCD monitor and the printed test images were viewed in a D50 light booth. The experimental results showed that the lightness, chroma, and hue differences behaved differently when the perceptual color differences were plotted against the computed differences. This implied that the color-difference formulas should be optimized and that different weighting factors should be applied to different visual attributes. The color-difference formulas can be optimized from the experimental data by the slope ratio of best-fit lines of lightness, chroma, and hue. The optimized formulas CIELAB(1.5:1), CIEDE2000(2.3:1), CIE94(3.0:1), and CMC(3.4:1) for digital images, or formulas CIELAB(2.4:1.5:1), CIEDE2000(2.8:1.6:1), CIE94(2.9:1.6:1), and CMC(2.7:1.5:1) for printed images, when considering hue, performed much better than the original formulas.

Abstract An accurate colorimetric characterization of digital still cameras (DSCs) is vital to any high-quality color-reproduction system. However, achieving a perfect relationship between DSC responses and input spectral radiance is not practically easy, even when they have a reasonable linear relationship. In this research, we investigated differences in capturing geometries as a source of nonlinearity in camera characterization workflows. This nonlinearity can be corrected using a physical model describing the spectrophotometric changes according to illumination/capturing geometries. We introduced a model based on the Saunderson equation as an approach to predict surface properties suitable for paint layers in different geometries. According to the results, the Saunderson surface correction successfully compensated for the dissimilarities among spectrophotometric and spectroradiometric measurements, regardless of the capturing and lighting geometries. The model was also used for characterizing digital still cameras using matte, semi-glossy and glossy color targets as training datasets. The Saunderson-based models improved the transformation matrix for different geometries compared to conventional methods. Also, the results confirmed the validity of a simpler derivation of the Saunderson surface correction based on linear matrix operations.

Abstract A blow-off tool has allowed for the measurement of the cumulative distribution of charged toner particle adhesion to a polymer belt. The tool methodology and design are described. The tool is calibrated to obtain the adhesion force using two different techniques, and the measured values agree well with published values using other adhesion measurement techniques. Measurements taken in different environments have yielded significantly different particle adhesions for the same average particle charge.

Abstract Both electrostatic and dispersive (van der Waals) forces contribute to particle adhesion, which has a significant effect on toner transfer in the electrophotographic process. Several approaches to adhesion measurements have yielded a large range of results for a variety of particle and environmental conditions. We present adhesion measurements taken in different environments using the metered air pulse method. They yield significantly different removal forces as a function of temperature for the same average particle charge. Particle deformation due to a combination of changes in particle stiffness with temperature and compressive electrostatic forces can predict the resulting adhesion increase. The morphology change is one of several factors which can contribute to the measured adhesion, which is significantly higher than values obtained by considering only the charged particle monopole and its image. Additionally, non-uniform charging in controlled adhesion experiments provides further muddling between the electrostatic and dispersive forces. This result is due to the electrostatic force having a component which is independent of the nominal charge under certain conditions. We find that the adhesion forces can be fully cubic with respect to the average particle charge, and that the components of the adhesion force may be much more difficult to decouple than previously thought.

Abstract An alternative low-cost replacement for silver and gold conductive inks is of great interest to the printed electronics industry. Nanoparticle copper inks and silver-coated nano-copper inks are some of the alternative materials being tested for use, especially in applications where low-temperature flexible substrates are favored. Although the inkjettability of nano-copper ink and the influence on print quality has been reported, information regarding the relationship between the ink film thickness and the energy required for sintering by intensive pulse light is not yet understood. In this study, an inkjettable nano-copper ink was printed on PET (polyethyleneterephthalate) and glass, and the samples were sintered using bursts of high-intensity pulsed light. The amount of energy applied determined the degree of sintering among particles. The greater the number of sintered nanoparticles, the higher is the conductivity of the printed traces. A comparison of energy levels required for sintering on glass and PET in relationship to the ink film thickness is reported, and the thermal contribution of the substrate to the processing energy requirements of this ink is revealed.