Spatial sharpening of thematic mapper multispectral images with a higher resolution panchromatic band image

Abstract

A method for spatial resolution enhancement, or sharpening of three bands of a Landsat thematic mapper (TM) multispectral image by a higher resolution SPOT panchromatic (pan) image, is described. Visible band TM images having positive correlation with the geometrically coregistered pan image are sharpened with a reported multiresolution sharpening technique. Application of neural network estimation technique is described to improve sharpening of near-infrared bands specifically when there are reverse contrast patterns at respective edge boundaries of the pan and TM images. Unless corrections are made for reverse patterns there will be degraded sharpening. Thus multilayer feedforward neural networks are trained to adaptively estimate, at multiple resolutions, pan edge patterns that have the same contrast as the TM edge patterns. Using this estimated data with a reported multiresolution sharpening technique gives spatial resolution enhancement. Image examples illustrate opposite contrast edges of pan and near-infrared TM bands. Other examples show improved sharpening when neural network modified edge information is combined with the multiresolution sharpening methods. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.