Classification of patterns on high resolution SAR images

Synthetic Aperture Radar being an all weather adaptive and deeply penetrating, forms an inevitable part of all processes of investigation. Classifying different patterns like rivers, buildings, land areas, farm land etc has got prominent role in remote sensing applications, military applications etc and hence has been actively researched in recent years. This paper presents a novel approach for classifying high resolution SAR images. Image denoising is the first step in certain applications like classification problem, pattern matching etc. Here a modified Non Local Means filter method is used for denoising and also explores the possibility of using Artificial Neural Networks (ANN) for classifying different patterns on high resolution SAR images based on a fusion method. The proposed method uses the features of Local Binary Patterns (LBP), features in RGB color space and features in HSV color space. The experiments on high resolution SAR images obtained from Quickbird and Ikonos satellites shows that the proposed method outperforms the other widely used feature extracting methods in SAR image classification.