Technical Report: Independent Component Analysis strategies for extracting information from remote-sensed images

CESAR F. CAIAFA

2005-09-01/2005-12-23

Electrónica

Espacio

This document reports the research results obtained in the period Sept - Dec 2005 at ISTI, CNR, Pisa, ITALY under the ASI-CONAE Fellowship Program.Title: Independent Component Analysis strategies for extracting information from remote-sensed images.Aim: A support to the extraction of information from images coming from remote sensing platforms.Context: The next generation of earth observation satellites will yield images with high spatial, spectral and temporal resolution. In order to separate homogeneous regions from panchromatic or narrowband images, we should rely on morphology or texture analysis. The expected availability of remote-sensed data with sufficient spatial, spectral and temporal resolution, however, allows us to devise extraction or separation techniques based on the spectral signatures of the different components. In principle, this could at least offer two advantages:    1) The homogeneous components to be extracted do not need to be spatially separated.    2) The features to be extracted do not need to be known in advance: the thematic information can be extracted on the basis of normally available additional knowledge.    The availability of high-resolution data, even though they come from different sensors or platforms, can highly favor spatial registration when needed. "Blind" analysis is independent of the features to be extracted from the data sets, and thus, depending on the kind of images that will be available, it could also be useful to assess the performance of the techniques developed in separating any kind of statistically distinct patterns.