Independent component analysis assisted unsupervised multispectral classification

The goal of unsupervised multispectral classification is to precisely identify objects in a scene by incorporating the complementary information available in spatially registered multispectral images. If the channels are less noisy and are as complementary as possible, the performance of the unsupervised classifier will improve. The discriminatory power of the classifier also increases if the individual channels have good contrast. Hence there is a need to preprocess the channels so that they have high contrast and are as complementary as possible. Independent Component Analysis (ICA) is a signal processing technique that expresses a set of random variables as linear combinations of statistically independent components. Performing ICA on the multispectral images yields images that have high contrast and are maximally complementary. Unsupervised classification on these images captures more information than on the original images. This paper presents some results based on this simple preprocessing step. In preliminary studies, using MR images of the brain, we were able to classify detailed neuro anatomical structures such as the putamen and choroid plexus, from the independent component channels. These structures could not be delineated from the original images using the same classifier.