Showing papers presented at "Color Imaging Conference in 1994"

A Comparison of Methods of Sensor Spectral Sensitivity Estimation.

Abstract: An experiment is described to obtain estimates of the spectral sensitivities of a digital image capture device by photographing a wide range of input color signals. Given the red, green, and blue signals from the device and the spectral power distributions of the color signals, we apply the methods of simple linear regression, rankdeficient pseudo-inverse, and Wiener estimation. Nominal data and the results obtained by the methods are presented and compared.

Visual Evaluation and Evolution of the RLAB Color Space.

Abstract: The RLAB color space has been tested for a variety of viewing conditions and stimulus types. These tests have shown that RLAB performs well for complex stimuli and not-so-well for simple stimuli. This paper reviews the various psychophysical results, interprets their differences, and describes evolutionary enhancements to the RLAB model that simplify it while improving its performance.

Realization of Color Constancy Using the Dichromatic Reflection Model.

Abstract: A method is proposed for recovering both the surfacespectral reflectance function and the illuminant spectralpower distribution based on the dichromatic reflection model. A multi-band imaging device is realized by combining a monochrome CCD camera and several color filters, to separate the incident light into six wavelength bands throughout the visible spectrum. We show a measuring method to predict the spectral sensitivity functions of the imaging system. The dichromatic reflection for inhomogeneous objects is described using finite-dimensional models. The estimation algorithms are presented to recover the scene parameters in two steps from the image data. The illuminant spectrum of a projector lamp is recovered fairly well, and the Machbeth color patches are estimated in the average accuracy of about ∆Eab = 3.5.

Optimal filters for the reconstruction and discrimination of reflectance curves

Abstract: Under the assumption that the interaction between objects and light is completely described by reflectance curves, we were able to determine filters to reconstruct and discriminate objects optimally. We found that in most cases the colour matching functions are not optimal and that sometimes better results are obtained with one well-chosen filter.6

Scene Normalization Mechanisms in Humans.

Abstract: Using a variety of complex displays these experiments compare human image processing with photographic, hybrid and electronic imaging systems. These experiments measure human responsiveness to changes in averages and maxima, as well as test for evidence of long-distance spatial interactions. The result of this comparison is that imaging systems would have improved color and tone-scale performance if they mimicked human vision.

When to Use Linear Models for Color Calibration.

Abstract: With the advent of vector-space approaches, linear estimation techniques can be used in various ways for different imaging scenarios. We compare two methods for constructing linear models for surface reflectance spectra, the PrincipalComponents Analysis (PCA) and the One-Mode Analysis (OMA) [1] applied in simulating image capture under a number of realistic lighting conditions. We demonstrate that successfully using such methods depends on the exact problem at hand.

Device-Independent Color via Spectral Sharpening

Abstract: Color sensors in scanners and color copiers are not colorimetric—RGB values are not a linear transformation away from device-independent XYZ tristimulus values. For a given set of targets or dyes one can readily find a best linear transform or use interpolation. However, when the possible targets are unknown, a data-independent transform is needed. Here, we set out a very simple linear transform for forming XYZ from RGB, developed in analogy with a wellknown solution for the color constancy problem in computer vision, based on using narrow-band sensors. In a scanner, we know the illuminant. Therefore the color constancy paradigm— illumination-independent colors—is not applicable. Instead, we change filters—from RGB to XYZ. The von Kries adaptation form of the color constancy solution can then apply if we can “sharpen” both the RGB sensors and the XYZ color-matching functions. Recently, we developed just such a “sharpening” basis transform: most of the sensitivity of the new possibly partly negative sensors is isolated in a particular wavelength interval. Here we “sharpen” both sensor sets; after dividing by sharpened white-spot values an inverse transform results in recovered XYZ values. Applying the method to 462 Munsell chips yields a median CIELAB error of only 3 units for two different systems.

Adjusting Simultaneous Contrast Effect for Dynamic Information Display.

Abstract: Simultaneous contrast effect often confuses the viewer’s understanding of the color coding scheme on information graphics, such as maps and diagrams. In particular, on a computer-based dynamic display, such as weather and air traffic, since background color and position of a graphical element are difficult to predict at run-time, we need to adjust the physical color of each element automatically so that all elements that are intended to appear the same color are perceived that way. This paper introduces experimental information graphics that automatically adjusts color differences based on Jameson and Hurvich’s research The results suggest that adjustment of simultaneous contrast effect can increase the reliability of dynamic information display and the flexibility of its design.

Colorimetric Characterization of a Desktop Drum Scanner via Image Modeling.

Abstract: A desktop scanner was colorimetrically characterized to average CIELAB error of less than unity for Kodak Ektachrome transparencies and Ektacolor paper, and Fuji Photo Film Fujichrome transparencies and Fujicolor paper. Independent verification on spectrally similar materials yielded average ∆E*ab error of less than 2.1. The technique first modeled the image formation of each medium using either Beer-Bouger or Kubelka-Munk theories. Scanner digital values were then empirically related to dye concentrations. From these estimated dye concentrations, either spectral transmittance or spectral reflectance factor was calculated from an a priori spectral analysis of each medium. The spectral estimates can be used to calculate tristimulus values for any illuminant and observer of interest. Results and Discussion The ∆E*ab histogram for the Ektachrome IT8.7/1 modeling target (as a representative material) is shown in fig. 1 where the predicted CIELAB values are compared with the measured CIELAB values. The histogram of an independent target with 216 patches (6x6x6 digital factorial design) is shown in fig. 2. The table lists the results for the four films analyzed. The 95th percentile was slightly greater than two times the mean of each data set. This technique was very effective for positive films and slightly less effective for photographic papers. In the case of Ektacolor Plus paper, the empirical model relating linearized scanner values (Dr, Dg, Db) to dye concentration had greater residual error than the transparent materials; this was due to the wide-band scanner spectral sensitivities and quantization errors at high concentrations. For the Fujicolor paper, the independent data had large systematic errors. In a post hoc analysis, it was found that the independent target had different dye characteristics than the IT8.7/2 target; as a result, the spectral reconstruction was in error.

Communicating Color Appearance with the ICC Profile Format.

Abstract: The advantage offered by the ICC system over proprietary color management systems is its potential interoperability. This not only requires agreement on a file format but also on the way this format communicates appearance. The successful use of the format requires both cross-media and cross-environment transformations. To achieve interoperability, the ICC format places these transformations in the profile rather than in the color management engine. This paper does not attempt to summarize the ICC specification. Instead, it establishes a broader context for understanding the way the design of the ICC communicates color.

Color Appearance in Images Measurements and Musings

Abstract: We suggest a framework for predicting color appearance of image data. The image data are converted by a series of standardizing transformations into a simpler stimulus with the same color appearance. By working through these standardizing transformations, we can build color appearance models for images without abandoning—or repeating—much of the work that has gone into defining the visual properties of the CIE standard observer.