Object oriented classification of high spatial imagery using Gabor textures and composite kernels

High spatial resolution images are widely used in many different areas. In order to increase classification accuracy, texture feature are widely studied. This paper presents a method of object oriented texture analysis method based on Gabor filters. In fusion of spectral features and Gabor textures, composite kernel methods were applied. The results show its effectiveness in extracting texture information using Gabor filters. And it is demonstrated that fusion of Gabor texture and spectral features using composite kernels is effective.