Synthetic aperture radar (SAR) images classification using speckle filtering and texture information

Synthetic aperture radar (SAR) is very efficient instrument for obtaining remotely sensed images of the earth's surface. However, SAR image is degraded by a form of multiplicative noise known as speckle, which is a result of the illumination, by the coherent radar. Hence speckle reduction is a necessary procedure before automatic image classification can be performed. This paper deals with the supervised classification of SAR images. Our approach consists in the speckle filtering before the clustering. But the only knowing of the filtered intensity image is not sufficient because of the high noise level. The other possible information to help the clustering is the texture, thus our approach is based on these two criteria.