New techniques for classification, enhancement, and fusion of multispectral remote sensing images

ABSTRACTMSc THESISNEW TECHNIQUES FOR CLASSIFICATION, ENHANCEMENT, AND FUSION OFMULTISPECTRAL REMOTE SENSING IMAGESGülçin YİĞİTLERDEPARTMENT OF ELECTRICAL AND ELECTRONICS ENGINEERINGINSTITUTE OF NATURAL AND APPLIED SCIENCESUNIVERSITY OF ÇUKUROVASupervisor : Asst. Prof. Dr. Turgay İBRİKÇİProf. Dr. Okan ERSOYYear: 2005, Pages: 69Jury : Asst. Prof. Dr. Turgay İBRİKÇİProf. Dr. Vedat PEŞTEMALCIAsst. Prof. Dr. Kerem ÜNRaw satellite images often contain a vast amount of information that is not readilyapparent to the analyst. Therefore, image enhancement techniques are used to highlightfeatures of interest and expose subtle differences in the spectral signature of the componentsof the target.This thesis presents a new approach to the development of multispectral /hyperspectral image enhancement and fusion algorithms. In approaches used up to now forimage enhancement, the bands are typically processed separately, and this results inconsiderable distortion. The amounts of information in many bands are also not veryefficiently used. The objective of this work is to utilize the amount of information availablemore effectively, remove such distortions and to improve the appearance of the images. Torealize this goal, we developed two new algorithms using transform techniques. In theresulting algorithms developed, image enhancement and image fusion are consideredtogether. The first algorithm developed is RGB (Red Green Blue) to IHS (Intensity HueSaturation) (RGB2NTSC) algorithm which uses RGB to IHS transformation method and amorphological filter. The second algorithm developed is the real discrete Fourier transform(RDFT) algorithm. Because the RGB2NTSC algorithm allows a transformation to work onlywith three images, to be able to combine more than three images, we used the RDFT. Thefinal images obtained from these algorithms are much less noisy and has more desirableproperties such as sharp edges.Keywords: Image Enhancement, Image Fusion, Multispectral and Hyperspectral Images, IHSTransform, RDFT Transform.