Related Experiment Videos
Linear models of surface and illuminant spectra.
1Xerox Palo Alto Research Center, California 94304.
Summary
Researchers developed new spectral representations for color, improving upon traditional methods. These efficient models enhance color accuracy in applications like computer graphics and printer calibration.
Area of Science:
- Color Science
- Computer Vision
- Image Processing
Background:
- Conventional color representations rely on human peripheral vision (tristimulus values).
- Approximating spectral properties of surfaces and illuminants is crucial for accurate color reproduction.
Purpose of the Study:
- To create efficient spectral representations for color that generalize conventional methods.
- To develop low-dimensional linear models for approximating spectral properties.
Main Methods:
- Utilized low-dimensional linear models to approximate spectral properties of surfaces and illuminants.
- Optimized linear-model basis functions by minimizing sensor response approximation errors.
Main Results:
- Developed generalized spectral representations outperforming principal-component approximations.
- Demonstrated conceptual simplifications for applications like printer calibration.
Conclusions:
- The proposed linear models offer efficient and accurate spectral representations for color.
- These models provide significant advantages over traditional and principal-component methods in computer graphics and calibration.