Texture Analysis for Classification of SAR Images

Abstract

SAR (Synthetic Aperture Radar) image classification has been a constant field of research since long. SAR image classification has numerous applications like map updating, oil spill detection in oceans, automatic target recognition etc. A lot of work has been done on use of Texture features, frequency spectrum features etc. for SAR image classification. But still there is scope for a simple but effective classification technique. In the present work, we have proposed a classification system for SAR images based on the highly discriminative texture features. We have studied 30 SAR images for 20 texture features. Statistical Approach has been used for texture analysis of the images. Finally, we have reduced the required number of texture features to a lesser number of features. The results being obtained are tested on test images to give an accuracy of 95% for image classification.