Illumination Invariant Color Representation
The goal of this project is to develop a color space transformation technique that adapts the color space to the color of the illuminant and leads a new color representation that is more independent of the illumination intensity and is more adapted to the spectral composition of the illumination than any of the existing approaches. The central notion of our new color space transformation scheme is that it aligns the vertical axis of the new color space with the color of the illuminant in the original RGB color space. This provides us two benefits: 1) Locating the vertical axis of the new color space in the direction of the brightness of the illuminant color gives us the adaptation need to achieve color constancy; and 2) The project of the pixel data onto the horizontal plane in the new color space provides us with a single scalar descriptor for the chromaticity -- as opposed to the two such descriptors that would be needed otherwise with two-dimensional color subspaces, such as the HS color subspace and the rg-chromaticity space.
    •    Jae Byung Park
 Jae B. Park and Avinash C. Kak, "A Truncated Least Squares Approach to the Detection of Specular Highlights in Color Images," Proceedings of the IEEE International Conference on Robotics and Automation (ICRA 2003), Taipei, Taiwan, 2003  [pdf, 0.5MB]
 Jae B. Park, "Efficient Color Representation for Image Segmentation under Non-white Illumination Conditions," Proceedings of the International Symposium on Photonics; Technologies for Robotics, Automation and Manufacturing, Providence, Rhode Island, U.S.A., 2003  [pdf, 0.5MB]