An important objective of visual neurophysiology is to understand visual perception in terms of neuronal connections. In the case of color vision, some of the chief aspects of perception to be explained are color mixing and color constancy. Color mixing is now well understood in physiological terms; progress has been made in the area of color constancy, but there our understanding is far from complete.Anatomically our visual system consists, broadly of a series of neuronal stages; the first three of these are housed in the retina, beginning with the rods and cones and ending with retinal ganglion cells, whose axons make up the optic nerve. Each ganglion cell receives input, via bipolar cells, from a small compact aggregate of rods and cones, which constitute the receptive field of that cell. In 1950, Stephen Kuffler, working with cats, found that each retinal ganglion cell receptive field is a few mm in diameter and is subdivided into a small excitatory (“on”) center and larger inhibitory (“off”) surround or, for about half the cells, an inhibitory center and an excitatory surround. This organization gives an immediate and simple explanation for the fact that object's whiteness or blackness depends on spatial differences in brightness, and not on the absolute levels of light coming from the object.

In this paperwe present a multiscale color appearance model which simulates luminance, pattern and color processing of the human visual system to accurately predict the color appearance attributes of spectral stimuli in complex surroundings under a wide range of illumination and viewing conditions.

A simple, uniform color space (the IPT color space) has been derived that accurately models constant perceived hue. The model accurately predicts hue without detrimentally affecting other color appearance attributes. Several psychophysical data sets have been modeled in the new color space and appear to perform as well as or better than CIELAB and CIECAM97s color spaces. Data sets tested and compared to CIELAB and CIECAM97s include Munsell renotation colors at Value 5, OSA color system uniform scale data, MacAdam's (observer PGN) equi-luminant color tolerance ellipses, suprathreshold color-difference ellipsoids (RIT-DuPont visual color difference data), lightness of chromatic object colors (Helmholtz-Kohlrausch effect), and the two constant hue data sets. Quantitative analysis is discussed for the constant hue data sets and the Helmholtz-Kohlrausch effect data. A verification experiment that compares the new space to Hung and Berns', and Ebner and Fairchild's constant hue data sets has been performed. Results show that the new space is judged to be at least as uniform as table based hue corrections derived from the data sets.

The CIE Board of Administration has proposed a new Division to study procedures and prepare guides and standards for the optical, visual and metrological aspects of the communication, processing, and reproduction of images, using all types of analogue and digital imaging devices, storage media and imaging media. While current practice enables adequate color communication within single industries, such as publishing or television, it was felt that more effort was needed to make the data produced within these industries readily useable in other areas. The CIE is the most appropriate body in which to do this since illumination, vision and measurement are all topics within its scope and are all essential components in specifying and controlling the digital reproduction of color images.

The recently developed CIE color appearance model, CIECAM97s, provides an extension of the previously recommended CIE color spaces. This paper examines the new color appearance model by comparing it to CIELAB, one of the more widely used color spaces for digital color imaging. First, the perceptual attributes common to both spaces are compared and contrasted. Second, examining device gamuts in CIELAB and CIECAM97s further highlights differences in the spaces. Lastly, the trends in the color differences for the two color spaces are assessed. The focus of this paper is to highlight differences using simple numeric calculations.

A set of psychophysical experiments was conducted to investigate backgrounds for determining the adapted white points of CRTs viewed under variously illuminated environments. A number of background characteristics were modified, such as pixel size, chroma range and lightness range. All backgrounds tested averaged to the same luminance and chromaticity. Observers viewed solid-colored samples on a field of each background displayed on a D65 balanced CRT monitor in a dark environment. Their task was to select the most achromatic appearing samples through an iterative process. Only two of the six backgrounds were found to result in near complete adaptation to the monitor: the control, a solid field of L*=60, and, an achromatic random dot pattern. None of the other tested backgrounds, which were all chromatic random dot patterns, resulted in complete adaptation, and all had very large variances. The conclusion is drawn that chromatic random backgrounds can significantly effect chromatic adaptation. This is true even if the measured background is neutral and if the background pixels average to an achromatic.

We measured the threshold of visibility of a thin line against a solid background. The line colour differed from that of the background in only one of L*, C*, and h_ab. The twenty-seven background colours were distributed throughout colour space. We observed trends in the dependence of the threshold differences in C* and h_ab as functions of C* and h_ab of the background colour. We found our best quantitative agreement resulted from a C* and h_ab dependent colour difference formula, similar in spirit to the CIE94 colour difference metric, but with different parameters and weighting functions. Finally we simulated our geometry in input to an s-CIELAB implementation and found that s-CIELAB predicted our results quite well.

In modeling color vision, certain visiblewavelengths have special significance. A growing body of scientific work shows that the wavelengths around 450nm, 540nm and 605nm, the so called prime-color (PC) wavelengths, are fundamental to color vision. Perhaps unsurprisingly, these same wavelengths are often discussed in the color imaging literature. Monitors that can display a large gamut of colors and are visually efficient have phosphor-primary peaks at the PC wavelengths. Color cameras that have peak sensitivities at the PC wavelengths have favorable color-balancing properties. Why are the PC wavelengths so important? This paper provides a start toward a mathematical theory to answer this question.

Sensor sharpening has been proposed as a method for improving color constancy algorithms but it has not been tested in the context of real color constancy algorithms. In this paper we test sensor sharpening as a method for improving color constancy algorithms in the case of three different cameras, the human cone sensitivity estimates, and the XYZ response curves. We find that when the sensors are already relatively sharp, sensor sharpening does not offer much improvement and can have a detrimental effect. However, when the sensors are less sharp, sharpening can have a substantive positive effect. The degree of improvement is heavily dependent on the particular color constancy algorithm. Thus we conclude that using sensor sharpening for improving color constancy can offer a significant benefit, but its use needs to be evaluated with respect to both the sensors and the algorithm.

It is commonplace to use a 3 × 3 linear transform to map device RGBs to XYZs. Two particular types of transforms have been developed based on the assumptions that we either maximally ignorant or maximally prescient about the world. Under the maximum ignorance assumption, it is assumed that nothing is known about the spectral statistics of the world and so the best correction transform is the one that maps device spectral sensitivities so they are as close to observer sensitivities as possible. Under maximum prescience, we know the spectral statistics that we will observe and so the maximally prescient transform maps, with minimumerror, the RGBs (that we knowwe will see) onto corresponding XYZs. In general the two assumptions lead to quite different color corrections.In previous work we have argued against total ignorance or prescience and have instead developed compromise transforms. Our work is based on two observations. First, one is never completely ignorant about the world—color signal spectral power distributions are everywhere all positive. Second, it is accepted that it is much more important to correct some colors than other. In particular, white is central to color vision and color imaging, so it is imperative that white should always look right.However, to date these two compromise solutions have been studied in isolation. Surely, it would be advantageous to combine the constraints of whiteness and positivity? In fact we show that this is not the case: by preserving white we enforce positivity. This is an important result. Not only does it add to our understanding of color correction, but it helps explain color correction results published in the literature (the assumptions of positivity and white-preservation lead to very similar results). Moreover, it helps us to derive a new measure for assessing the goodness (color correctability) of camera sensors that is strictly less pessimistic (and more accurate) than the existing Vora Value.