Mining very high resolution complex-valued SAR images using the fractional Fourier transform

In this article, we present a new spectral descriptor for complex-valued SAR images for information mining in a large repository of images. This descriptor is based on the fractional Fourier transform (FrFT) computed at different transform order for an image patch. FrFT can be considered as a decomposition of signals in terms of chirps. The method is demonstrated for TerraSAR-X High Resolution Spotlight SLC images. Performance evaluation is carried out by comparative study of proposed method with descriptors obtained from Grey-Level Co-occurrence Matrix (GLCM) of the patch. With the presented results, we emphasize the use of complex-valued SAR images for target characterization.