Abstract

Color image processing is often done by applying monochrome image processing techniques to the individual spectral components. However, such an approach is often sub-optimal and neglects cross-spectral correlations. An alternative approach is to treat the image as a vector-valued function and use vector-based algorithms in a 3-D vector space. This can lead to improved enhancement, analysis, and visualization.

© 1993 Optical Society of America