A Novel Approach To Extract Road From Speckle Noise Corrupted Satellite Images Using Anisotropic Diffusion Filter And Morphology

Abstract

Extraction of road from satellite images plays a major role in navigation, monitoring roads, traffic management and urban planning [1]. Satellite images are generally degraded by Salt and Pepper noise, Gaussian, Speckle noise, etc. Speckle noise is specifically observed in images recorded by Synthetic Aperture Radar [2]. Speckle noise arises during image acquisition due to backscattered signals from multiple targets on surface of the earth and hence it affects image interpretation. In this paper, anisotropic diffusion is used to remove speckle noise which also preserves the edges and other significant details in the satellite image. Thereafter, adaptive thresholding, morphological operations and edge detection are used to finally extract the desired roads.