SAR Image Segmentation Based on Multiresolution GLCP in Overcomplete Brushlet Domain

Based on the multi-scale direction characteristics of overcomplete brushlet transform, a feature extraction method called multiresolution grey level co-occurrence probabilities (GLCP) in overcomplete brushlet domain is presented, which makes full use of the direction information in different subbands. The segmentation results of Brodatz mosaics and synthetic aperture radar (SAR) image show that the proposed feature extraction method outperforms other methods such as brushlet, GLCP and DWT in the segmentation accuracy on the synthetic mosaic and synthetic aperture radar (SAR) image