A method of channel reduction is presented which reduces the amount of data to be processed during spatial image processing. Using a modified principal component analysis on non-overlapping blocks of an image, the spatial and spectral information in an image can be separated with very little loss. The processing need be applied only to the spatial information and the color information added after processing.
Hans B. Wach, Edward R. Dowski, W. Thomas Cathey, "Channel Reduction and Applications to Image Processing" in Proc. IS&T 7th Color and Imaging Conf., 1999, pp 317 - 320, https://doi.org/10.2352/CIC.1999.7.1.art00061