Bag-of-Visual-Words Model for Classification of Interferometric SAR Images

This work introduces a well-accepted image representation model in image analysis, namely the Bag-of-Visual-Words (BoVW), to interferometric SAR (InSAR) images. As the low-level local features, Gabor- and fractional Fourier transform (FrFT)-based feature descriptors are used. The supervised classification results with BoVW-Gabor and BoVW-FrFT features are compared to those with global Gabor and global FrFT features. Although the global Gabor features are better than the global FrFT features, by the implementation of BoVW model, FrFT outperforms Gabor features. Also, the classification performances of different baseline acquisitions for the same scenes are compared. For each baseline, the mean and individual class accuracies are improved by using BoVW- FrFT features.