Multivariate Contrast Enhancement of Color & Multispectral Images

Students: Phillip A. Mlsna, Qiang Zhang

Multispectral and true-color images are often enhanced using histogram-based methods, usually by adjustment of color components after transformation to a selected secondary color system. Enhancement aimed toward the preservation of certain important perceptual qualities generally calls for the secondary coordinate system to be perceptually based. However, independent modification of the secondary components seldom uses the full extent of the RGB gamut unless some color values are clipped at the RGB boundaries. Preserving perceptual attributes is sometimes less important than obtaining the greatest possible color contrast improvement. This is especially true for color composites derived from multispectral images, which have no significant basis in human perception. We developed new, multivariate enhancement techniques, such as "histogram explosion," which are able to exploit nearly the full RGB extent without clipping. While not generally based upon a perceptual model, these methods can preserve original hue values when parameters are chosen properly. Experimental results demonstrated that dramatic improvement in the color contrast can be achieved through these multivariate histogram modification techniques.