A Residential Area Extraction Method for High Resolution Remote Sensing Imagery by Using Visual Saliency and Perceptual Organization

Inspired by human visual cognitive mechanism, a method of residential area extraction from high-resolution remote sensing images was proposed based on visual saliency and perceptual organization. Firstly, the data field theory of cognitive physics was introduced to model the visual saliency and the candidate residential areas were produced by adaptive thresholding. Then, the exact residential areas were obtained and refined by perceptual organization based on the high-frequency features of multi-scale wavelet transform. Finally, the validity of the proposed method was verified by experiments conducted on ZY-3 and Quickbird image data sets.