2007 A comparison of radiometric normalization methods when filling cloud gaps in Landsat imagery

Mapping persistently cloudy tropical landscapes with optical satellite imagery usually requires assembling the clear imagery from several dates.  This study compares methods for normalizing image data when filling cloud gaps in Landsat imagery with imagery from other dates.  Over a complex tropical island landscape, St. Kitts/Nevis and St. Eustatius , all of the methods tested reduce inter-date image differences for ETM+ bands 1-5, 7, NDVI and the band 4:5 ratio.  Regression tree normalization reduces the inter-date differences more consistently than linear regression or histogram matching.  Normalizing ETM+ images with regression trees can produce more seamless imagery than linear normalization, histogram matching, or image-based atmospheric correction via dark object subtraction.  More seamless imagery enhances visual interpretation and helps reveal the distributions of forest formations in these landscapes.  Decision tree classification of cloud-filled Landsat imagery can accurately map land cover and detailed forest formations.  Classification accuracy is not highly dependent on the method used to make the cloud-filled imagery, however, at least as long as: 1) classification model training data reflect class spectral variability, and 2) ancillary spatial data that relate to the distributions of classes are used in the classification. Cloud-filled imagery is also known as cloud-cleared imagery.  The article was published in the Canadian Journal of Remote Sensing (http://pubs.nrc-cnrc.gc.ca/cjrs/cjrs.html).