Performance Comparison of Texture based Approach for Identification of Regions in Satellite Image

Human vision is the most important resource of information used for object recognition and classification. Images having constant intensities can be easily represented by vision. Textures are one of the important features in computer vision as it identifies different regions of an image on the basis of texture properties. It is widely used in variety of applications. Identifying various regions in satellite image is one such application. There are numerous approaches based on texture classification that are mainly categorized as statistical, structural, model based and signal processing methods. The study involves the classification of LANDSAT ETM+ and MODIS satellite imagery datasets using texture based approaches i.e. Grey Level Co-occurrence Matrices (GLCM), Laws Energy Measure, Discrete Fourier Transform (DFT) and Gabor Filter. Relative performance comparison study of these approaches on the basis of standard deviation (statistical tool) has been carried out. GLCM shows best results among all other approaches.