Wavelet projection pursuit for feature extraction and cloud detection in AVIRIS and AVHRR imagery

We examine a class of constrained projection pursuit (PP) algorithms for extracting textural features from multi-spectral remote sensing imagery. Based on the assumption that spatial frequency information is useful for separating classes of interest in the data, topological constraints are defined for the PP filter vectors. The constraint on each filter is imposed by a set of tunable meta-parameters which define each filter as an adaptive Gabor wavelet. We call this approach wavelet projection pursuit (WPP). The application of the approach to cloud detection is described. The long-term goal is to develop algorithms for texture-based cloud masking applicable to future data from the Multi-Angle Imaging Spectrometer (MISR).