Consider how imaging is used commonly by various professionals. Photographers are just as likely to use a digital camera as one loaded with film. Although they expect the images to look different, they may not know why they do. If they want them to match, they probably don't know how, or they won't know why they might never exactly match. A graphic designer works with direct-capture digital-image data, scanned photography, and synthetically rendered imagery. (The same is true for a movie editor.) In the long run, it is not cost effective to visually edit each image iteratively until they appear to originate from a single imaging device. A computer scientist may be expected to understand all computer peripherals, including digital-imaging devices such as scanners, displays, and printers. They may need to deal with setting up the color preferences in Photoshop, creating ICC color profiles, and using various hardware and software to achieve acceptable color quality. A librarian is expected to handle digital image archives as seamlessly as a collection of books. An art historian viewing a downloaded image from a museum website (or any image database) assumes that their perceptions are similar to viewing the actual work of art. An archivist is expected to digitize photographic reproductions such that the digital archive is an accurate color reproduction of the original work of art. Professionals dealing with digital imaging are suddenly expected to have the expertise of knowledgeable color imaging scientists and engineers. (Many of these professionals may not even realize that there is such a thing as an imaging or color scientist.) These practices, assumptions, and expectations are real. Are they realistic? Are they achievable? To a large extent, they are realistic and achievable by treating imaging as an analytical tool, a scientific tool. How can typical imaging practices become scientific imaging practices? This is the subject of this keynote address. The following will be discussed: Education in imaging and color science, Review of the human visual system, Input device spectral sensitivity, bits and encoding, color management, image quality metrics, characterization and calibration targets, multi-channel visible spectrum imaging, metadata, and standards.

A fundamental problem in psychophysical experiments is that significant conclusions are hard to draw due to the complex experimental environment necessary to examine color constancy. An alternative approach to reveal the mechanisms involved in color constancy is by modeling the physical process of spectral image formation. In this paper, we aim at a physical basis for color constancy rather than a psychophysical one.By considering spatial and spectral derivatives of the Lambertian image formation model, object reflectance properties are derived independent of the spectral energy distribution of the illuminant. Gaussian spectral and spatial probes are used to estimate the proposed differential invariant. Knowledge about the spectral power distribution of the illuminant is not required for the proposed invariant.The physical approach to color constancy offered in the paper confirms relational color constancy as a first step in color constant vision systems. Hence, low-level mechanisms as color constant edge detection reported here may play an important role in front-end vision. The research presented raises the question whether the illuminant is estimated at all in pre-attentive vision.

In this paper, the performance of chromatic adaptation transforms based on stable color ratios is investigated. It was found that for three different sets of reflectance data, their performance was not statistically different from CMCCAT2000, when applying the chromatic adaptation transforms to Lam's corresponding color data set and using a perceptual error metric of CIE ΔE94. The sensors with the best color ratio stability are much sharper and more de-correlated than the CMCCAT2000 sensors, corresponding better to sensor responses found in other psychovisual studies. The new sensors also closely match those used by the sharp adaptation transform.

This paper proposes a chromatic adaptation model based on spectral property estimation. In this model, the concept of color constancy in human vision is introduced into color matching. We performed detailed subjective experiments to evaluate the color matching performance of the model for natural color images between softcopy and hardcopy by comparing it with that of six other models. The results not only show the color matching performance of each model, but also demonstrate that our model enables better color matching than the six other models, as seen in a darkroom experiment involving color matching between two CRT monitors whose whites are quite different.

Skin colour segmentation is important for human face tracking. An often used approach is to approximate the skin chromaticity distribution with a statistical model, e.g. with the distribution's covariance matrix. The advantage of this approach is that it is invariant to size and orientation and fast to compute. A disadvantage is that it is sensitive to changes of the illumination colour.This paper investigates how accurately the covariance matrix of facial skin chromaticity distributions might be modelled for different illumination colours using a physics-based approach. Results are presented using real image data taken under different illumination colours and from subjects with different shades of skin. The eigenvectors of the modelled and measured covariances deviate in orientation about 4o. This seems to be within a useful range for skin colour segmentation, and hence allow the statistical model to adapt to illumination changes.

It is well established that in order to obtain the best colour performance of a colour input device such as a scanner or a camera, that one needs to know the device spectral sensitivities. Unfortunately measuring sensitivities outside the laboratory is hard and moreover, manufacturers are reluctant to give the user specifications. Thus, there has been considerable interest in developing numerical techniques for estimating the spectral sensitivities.These methods are based on taking images of known spectral targets and then, using knowledge of the image formation process, solving for the sensitivities using numerical methods. It is important to state that while these methods perform reasonably well, the problem is inherently ill-posed. There is simply not enough degrees of freedom in the spectral profile of a reflectance target to recover device sensitivities.In this paper we tackle this uncertainty head on and develop a method to recover device sensitivities with uncertainty error bars. Experiments with a Megavision camera return a sensor estimate together with error bars. The error bars are sufficient to explain the discrepancy in the recoveries delivered by single-answer estimation algorithms and the actual sensitivities.

A novel biological model for color contrast is presented. An additional goal of the model was to achieve automatic color correction of still and video images. The model predicts human visual performance according to the physiology of the first and second order of the colorcoded cells in visual system. It is based on the properties of retinal ganglion cells (opponent cells) and cortical cells (double opponent cells) as well as on chromatic adaptation mechanisms in these double opponent colorcoded cells: remote chromatic adaptation. The simulations calculated the perceived image for still images, and were performed in order to correct image colors. The results indicate that the contribution of adaptation mechanisms to color contrast is significant, robust, and enables color correction of still images.

When a pair of surfaces is partially covered by a transparent filter, the ratios of cone excitations for the surfaces viewed directly are almost identical to those when the surfaces are viewed through the filter. We have previously shown that in simulations of Mondrian-like patterns partially covered by filters the invariance of the cone-excitation ratios predicts psychophysical performance in discrimination tasks. In this paper we investigate whether the number of surfaces in the display affects the strength of the transparency percept when observers are required to discriminate between displays with almost perfect invariance of cone-excitation ratios and similar displays where the ratios have been perturbed by noise. We find that discrimination performance increases with the increasing number of surfaces in the display. We also find that noise added to the S-cone class alone did not affect discrimination performance.

The currently used colour rendering index is based on an outdated colour space and chromatic adaptation formula. The perceived and calculated colour rendering of new light sources shows discrepancies. This becomes acute with the introduction of LED sources and increased use of coloured samples produced by ink-jet printers and colourphotocopy. Recent colour appearance models enable an advanced description of colour rendering. The present paper shows some visual experimental results and draws attention on possible updates of colour rendering calculation.

As part of the 5^th Framework Programme, the European Commission supports research and development activities in the field of digital imaging, in particular within the Information Society Technologies (IST) specific Programme. This paper briefly introduces the IST Programme and gives an overview of collaborative projects most relevant to imaging research. The EU's 6^th Framework Programme is also briefly discussed and orientations for the integration of imaging research are proposed.