DIGITAL IMAGE PROCESSING OF LANDSAT IMAGERY

Abstract

This study discusses first the basis of renote sensing and the various media useHn capturing and storing remotely sensed images. Since LANDSAT II MSS data is being used in this study, the LANDSAT MSS system and calibration was discussed. An attempt is made to reduce radiometric errors in the data that were caused by decimal truncation during NASA's initial calibration. The results for one particular method is encouraging but it is found to be effective only for lCM value pixels associated with water. Overland data shows little changes when re-calibration is applied. Density splicing method for translating a LIINDSAT digital image into a grey scale or grey density map and a systematic approach to create an effective grey scale with standard ASCII overprint on a line printer is discussed. Next, Band Ratio method of image enchancement is presented in detail in relation to reflectance value conversion. Its use for image enhancement through the manipulation of grey scale offset factor and density slicing factor is deliberated in detail. Based on the characteristics of spectral signature, a Band Differential micthod is proposed and implemented as a means to complement the Band Ratio method. Coputer software has been developed as as aid to the analysis of pixel clustering and distribution for Band Ratio and Band Differential methods of image enhancement. The vector connectedness 4-dimensional clustering algorithm is choosen as a basis for this study involving digital image classification. The thesis also discuss the physics of clustering vectors in a multidime.nsional field and how the principle of delineration of a cluster of vectors in an image leads to the delineration of the features they represented. A change is proposed for the 4-dimensional clustering algorithm to modify it for use on small images where statistical noise would interfere with clustering. The proposal has been implemented in the form of an interactive software package on a mini-computer.