A wavelet based algorithm for pan sharpening Landsat 7 imagery

At IGARSS 2000 the Data Fusion Committee began the sponsorship of a series of contests with which the committee aimed to reach two goals. The first goal was to assess the state of the art of a clearly defined aspect of the data fusion field by empirically evaluating algorithms. The second goal was to identify collective weaknesses of current algorithms, so as to identify requirements of further research. The first contest focused in the sharpening of 30-m resolution multispectral images by using 15-m panchromatic imagery. The objective of the sharpening was to improve the spatial resolution of the multispectral imagery, while preserving the spectral information in homogeneous areas. For the IGARSS 2000 contest, Landsat 7 Thematic Mapper data acquired over urban and agricultural areas was used. Researchers from the Remote Sensing Technologies Center at Mississippi State University won the contest. This paper describes the pan sharpening methodology used.

I. Introduction

The IGARSS 2000 data fusion contest focused on the sharpening of 30 m resolution multispectral images by using 15 m panchromatic imagery. The objective of the sharpening is to improve the spatial resolution of the multispectral imagery, while preserving the spectral information in homogeneous areas. For the IGARSS 2000 contest, Landsat 7 Thematic Mapper data acquired over urban and agricultural areas was used (see Figure 1). Specifically, the contest rules called for using the panchromatic band (520–900 nm) to sharpen bands 1 (450–515 nm), 2 (525–605 nm), and 3 (630–690 nm). The results were to be evaluated electronically using a set of test pixels both of homogenous regions (e.g., large agricultural fields, and water) and of specific structures (e.g., waterways and roads). The spectral signature of the homogenous fields were to be as close to the original values as possible; and the width of the linear structures should approach the width seen in the panchromatic data.