SAR patch scene categorization

This paper presents SAR image classification based on feature descriptors within the discrete wavelet transform (DWT) domain using non-parametric features. Each wavelet based subband was transformed using a Fourier transform in order to evaluate spectrum properties of wavelet subbands. The first and second moments, Kolmogorov Sinai entropy and coding gain, were used for the non-parametric features within an oriented dual tree complex wavelet transform (2D ODTℂWT). A database with 2000 images representing 20 different classes with 100 images per class was used for estimation of classification efficiency. A supervised learning stage was implemented with support vector machine using 10% and 20% of the test images per class. The experimental results showed that the non-parametric features achieved 94.3% accuracy, when 20% of database was used for supervised training.