Spatial Resolution Improvement by Merging MERIS–ETM Images for Coastal Water Monitoring

Abstract

The MediumResolution Imaging Spectrometer (MERIS) was launched in March 2002 and has been providing images since June 2002. Before its launch, we had implemented a method to improve its resolution by merging its images with Landsat Enhanced Thematic Mapper images in order to preserve the best characteristics of the two images (spatial, spectral, temporal). We now present the results of this method for real MeRIS images (level 1b and 2) in a coastal area. The robustness of the method is studied as well as the influence of the delay between the acquisitions of the two images . Index Terms − Fusion, Landsat Enhanced Thematic Mapper (ETM), Medium Resolution Imaging Spectrometer (MERIS).

Unmixing-Based Landsat TM and MERIS FR Data Fusion

“…This multitemporal data fusion exercise will be of great interest for land cover mapping and for monitoring vegetation dynamics (e.g., in terms of fraction of absorbed photosynthetically active radiation, leaf area index, or chlorophyll content) at high spatial, spectral, and temporal resolutions. Nevertheless, it is important to realize that these fused images will only be an approximation of what the MERIS sensor would be measuring if it had a spatial resolution of 25 m. In addition, possible landscape changes between the dates of the Landsat TM acquisition and the MERIS images might further affect the quality of the fused images [7], since the number and location of land cover classes may change if the time span becomes too wide.…”

“…Using k = image size, therefore, results in a fused image with a low spectral dynamic range, where each of the classes is represented by an approximation of its mean spectral response. The latter approach was the one used by Minghelli-Roman et al [6], [7]. Although it is computationally fast (we only need to solve one system of equations), here, we prefer to also optimize the size of the neighborhood k such that we can account for the natural variability of the components present in the scene.…”

“…As a result, the fused image has (potentially) more information than each of the original images. A simplified version of this data fusion approach has been used by Minghelli-Roman et al [6], [7] image, needs to be optimized, because they solved the unmixing equations for the whole image at once. A large number of classes (typically > 100) are needed to achieve good results with this method [6], [7].…”

“…Furthermore, solving the unmixing equations for the whole image at once might severely hamper the quality and usability of the fused images, because all pixels belonging to one class will get the same spectral signature. In other words, if we apply the method as described by Minghelli-Roman et al [6], [7], we implicitly reject all the within-class variability. Therefore, we believe that using a neighborhood should be preferred over simultaneously solving the unmixing for all the pixels present in the scene.…”

Sparse Spatio-spectral Representation for Hyperspectral Image Super-resolution

“…In spatio-spectral image fusion, one class of methods exploits unmixing ( [22], [35]) for improving the spatial resolution of the hyperspectral images. These methods only perform well for the cases when the spectral resolutions of the two images are not too different.…”