Speckle Reduction in SAR Images by Using Discrete Wavelet Transform
Saturnino Leguizamón · 1997
SAR (Synthetic Aperture Radar) images are affected by the presence of a multiplicative noise, called the speckle. This noise gives these sort of images the typical salt and pepper aspect. When this information is used in combination with optical images (e.g. TM, SPOT images) this feature of radar images is a drawback . Therefore, assuming some statistical properties of the speckle, different types of spatial filters are being used to overcome this problem. In the last decade Wavelet Transform (WT) has an increasing interest in the field of remote sensing due to its advantage in the analysis of non-stationary signals and other advantages in respect of Fourier analysis. In this paper we explore the application of WT to reduce the speckle. Wavelet transform provides a complete description of the image at all available scales by means of the multiresolution analysis (MRA). Using this technique, which is a pyramidal analysis, the image under study can be decomposed in lower resolution images representing the general structure of it, and images with high frequency features in which the speckle is included. The approach used here operates on the decomposed image by reducing or modifying some of the high frequency components and, after that, reconstructing the filtered new image. Some experiments has been conducted using this approach on different radar images. The preliminary results of these experiences show that this approach offers a promising future.