An advanced, automated method is presented for determining an effective scene illuminant chromaticity (scene illuminant plus imaging system variables) from specular highlights in digital images subsequent to image capture. Underlying theories are presented based on a two component reflection model where the scene illuminant relative spectral power distribution is preserved in the specular component. Related methodologies for extracting scene illuminant information as well as alternative methods for achieving color constancy are presented along with factors which inhibit successful implementation. Following, development of a more robust algorithm is discussed. This algorithm is based on locating the center of convergence of a radial line pattern in the two-dimensional chromaticity histogram which theoretically identifies the effective scene illuminant chromaticity. This is achieved by using a radiality index to quantify the relative correlation between a radial mask and the histogram radial line pattern at discrete chromaticity coordinates within a specified search region. The coordinates associated with the strongest radiality index are adopted to represent the effective scene illuminant chromaticity. For a set of controlled test images, the physics-based specular highlight algorithm determined effective scene illuminant chromaticities to a level of accuracy which was nearly three times better than that of a benchmark statistically-based gray-world algorithm. The primary advantage of the specular highlight algorithm was its sustained performance when presented with image conditions of dominant colors, weak specular reflections, and strong interreflections.
Library of Congress Subject Headings
Color printing--Quality control; Digital printing--Quality control; Color printing--Quality control--Data processing
- Please Select One -
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Richard, Wayne, "Automated detection of effective scene illuminant chromaticity from specular highlights in digital images" (1995). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus