A new multispectral system developed at the National Gallery is presented. The system is capable of measuring the spectral reflectance per pixel of a painting. These spectra are found to be almost as accurate as those recorded with a spectrophotometer; there is no need for any spectral reconstruction apart from a simple cubic interpolation between measured points. The procedure for recording spectra is described and the accuracy of the system is quantified. An example is presented of the use of the system to scan a painting of St. Mary Magdalene by Crivelli. The multispectral data are used in an attempt to identify some of the pigments found in the painting by comparison with a library of spectra obtained from reference pigments using the same system. In addition, it is shown that the multispectral data can be used to render a color image of the original under a chosen illuminant and that interband comparison can help to elucidate features of the painting, such as retouchings and underdrawing, that are not visible in trichromatic images.

The CRISATEL multispectral acquisition system is dedicated to the digital archiving of fine art paintings. It is composed of a dynamic lighting system and of a high resolution camera equipped with a CCD linear array, 13 interference filters and several built-in electronically controlled mechanisms. A custom calibration procedure has been designed and implemented. It allows us to select the parameters to be used for the raw image acquisition and to collect experimental data, which will be used in the post processing stage to correct the obtained multispectral images. Various techniques have been tested and compared in order to reconstruct the spectral reflectance curve of the painting surface imaged in each pixel. Realistic color rendering under any illuminant can then be obtained from this spectral reconstruction. The results obtained with the CRISATEL acquisition system and the associated multispectral image processing are shown on two art painting examples.

The analysis and synthesis of human skin color is important in many application areas. The skin color depends on some histological variables such as the concentration of pigments melanin and hemoglobin in skin layers. The present paper proposes a method for estimating the surface spectral reflectance of human skin based on an optics model and applying the estimates to 3D realistic image rendering for a human hand. The human skin is modeled as the two layers of the turbid materials for the epidermis and dermis. An estimation algorithm for the two layer model is then developed using the Kubelka–Munk theory. The parameters representing the concentration of pigments are determined based on spectral reflectance measurements of the human skin surface. In the application step, we describe a technique for rendering realistic skin images of a skin surface as a 3D object. The Torrance–Sparrow model is adapted in the image rendering process. The accuracy of the estimated reflectances is shown in experiments, and skin color images are created under a variety of illumination, viewing, and pigmentation conditions.

In this article we address the problem of performance of preprocessing before color image segmentation. The main goal of preprocessing is noise removal. Our interests are limited to nonlinear color filters working in the spatial domain. Most often comparing such filters is based on calculation of different quality factors (e.g. PSNR, NCD etc.). The main idea of this article is to use an evaluation function, coming from research on segmentation, to evaluate the performance of preprocessing. The experiments were realized using both original and noisy images corrupted by Gaussian and impulsive noise.

With the recent advances in color management systems, the estimation of accurate colorimetric values of objects being imaged by the use of sensor responses becomes an important function of color image acquisition devices such as digital cameras and color scanners. Therefore, colorimetric evaluation of a set of sensors is important to evaluate the colorimetric performance of the sensors. It is well known that the colorimetric quality of the sensors depends not only on the spectral sensitivities but also on the noise present in them. Although several evaluation models have been proposed, application of the models to real color image acquisition devices has not been appeared since it was impossible without prior knowledge about the noise present in the devices. In this article, a new model to estimate the noise variance of an image acquisition system is proposed based on the colorimetric evaluation model and for the first time the evaluation model was applied to real multispectral cameras by using the estimated noise variance. It is confirmed by the experiments that the proposed colorimetric quality is in good agreement with the experimental results and that the noise variance of the image acquisition system can be accurately estimated by the proposal.

Wide color gamut displays using monochromatic light sources, such as LED or laser, or applying multi-primary techniques have been introduced by several authors in forms of laboratory based models and for high end natural color vision applications. In order to bring these technologies into the consumer market, one of our latest investigations resulted in a rear projection type high definition multi-primary display (MPD) system which has been realized by a modification of our previous three channel DLP™ projection television. Thereby the full spectral energy of the UHP™ type projection lamp is sequentially separated into five primary spectrums through a rotating color wheel instead of the conventional RGB primary colors so that it enables to reproduce a wide range of colors than sRGB compatible display. The main focus of this study is to describe a gamut mapping method that allows using a full range of color gamut of MPD for a given limited color gamut of television signal. This approach is to display an image in such a way that the limited color saturations of the current television signal are adaptively enhanced for every observer. Through the gamut mapping, the picture quality of MPD is significantly improved in comparison with a conventional display system.

The camera capability to measure small color differences between sample pairs is evaluated by comparing the camera performance with a reference instrument. To this end, the appropriate working conditions are established, the camera spectral sensitivities and imaging noise are characterized, and the transformation to obtain a device independent representation of color is calculated considering two approaches: one, on the basis of the camera spectral sensitivity (CSS), and two, on the basis of the unified measure of goodness of the camera (UMG) that involves an imaging noise model. The camera performance is assessed from the measurement results of a large number of varied small color differences in the very pale and the dark grayish color regions, the involved uncertainty, the absolute discrepancy, and the relative discrepancy with respect to the reference instrument. In the experimental application, the three CCD camera SONY DX-9100P is assessed and compared with the spectroradiometer Photo Research PR-715 as reference instrument. The results reveal a high quality performance of the camera system, with absolute discrepancies in the estimation of color differences around the camera tolerances (CIELAB 0.5ΔE*_ab or CIEDE2000 0.6 ΔE₀₀). The color uniformity in textile dying is evaluated by analyzing some pairs of extreme center fabric samples. Although the camera is more sensitive to the texture effects than the spectroradiometer, both instruments yield consistent and satisfactory Pass/Fail results.

Color consistency is crucial for both photo and commercial printing applications. Dot gain tables are updated regularly, however between updates colors can shift due to process drift in the press, which is a common problem of both digital and offset presses. The goal of this investigation is to dynamically control the dot gain table and developer voltage to ensure more consistent color control while minimizing waste and calibration measurements. In this article we approach the elements of this calibration process as a series of machine-learning problems and investigate the efficacy of replacing physical calibration measurements with model-based predictions. The current state of the machine, expressed as sensor measurements, is used to model both the developer voltage, and the subsequent dot gain look up table. We also consider models that make a prediction based on a restricted set of calibration measurements, not necessarily including the full machine state vector. Our initial investigation using a preliminary dataset shows that machine learning methods are suitable for predicting the dot gain table.