Classification for Remote Sensing Images Based on Gabor Ring Filters and Ordered Spectral Histograms

Classification of remote sensing textures using Gabor ring filter and ordered spectral histogram is investigated. The proposed approach is based on three steps. First, multi-scale Gabor ring filters are designed for feature extraction. Second, marginal distributions, or called histograms, derived from filtered images by Gabor ring filter, would be sorted by their characterized values. Then we could get ordered spectral histogram which is the feature we proposed in this paper. Third, K-NN classifier using χ2 -statistic is applied to classify images. The proposed approach is applied in experiments on high-resolution remote sensing data. In our experiment, the proposed method performs well in terms of classification accuracy.