Wavelet-Fourier descriptors for SAR patch scene categorization

Abstract:
This paper presents SAR data categorization using discrete wavelet transform (DWT) and non-parametric features. The wavelet transformed subbands were additionally transformed using a Fourier transform and features within a Wavelet-Fourier domain were estimated. A supervised dictionary learning using a sparse representation was used for feature extraction. A database with 2000 images representing 20 different classes with 100 images per class was used for estimation of classification efficiency using supervised learning. The experimental results showed that the non-parametric features achieved 94.6 % accuracy, when 20 % of database was used for supervised training.