Influence of Number of Features on Texture Based Residential Area Extraction from Remotely Sensed Imagery

In this study, we implement and compare two texture analytical methods, 2-D Gabor and GLCM, for residential area extraction from high spatial resolution remotely sensed imagery. We elaborately investigate the influence of different settings to the outputs of both methods. Experiments show that both methods have the potential for texture based residential area segmentation, while 2-D Gabor presents an overall higher precision and consistency. Increasing the number of features, however, seems to contribute little to the accuracy of both methods.