Multiresolution filtering and segmentation of multispectral images

Abstract

We consider multiple resolution methods for filtering and segmenting multispectral astronomical images. For filtering, we use noise modeling, wavelet transform, and the Karhunen-Loeve transform. For segmentation, we use a quadtree followed by the fitting of a Markov model. We illustrate these methods on Hubble Space Telescope near infrared NICMOS camera images.