Textural information in SAR images for land-cover applications

Investigates the existence of textural information in spaceborne SAR images. ERS-1 SAR images processed at DLR (Germany) and NRSA (Hyderabad) are used for texture based analyses. The possibility of textural discrimination of different cover types such as paddy, sugarcane, water, urban areas, bush and shrubs are explored. Texture measures based on both the first and the second order image statistics are computed. The effectiveness of coefficient of variation (CV) as a texture measure is evaluated. The usefulness of gray-level co-occurrence matrices (GLCM) as a second order statistical measure of texture is investigated. The temporal plots of sample land cover categories using two texture features namely contrast and inverse difference moment (IDM), are used to qualitatively evaluate the separability of different cover types. Quantitative methods of separability, using two class separability measures are used to assess the usefulness of GLCM derived texture images. It is found that improvement in separability of some land-cover categories is obtained using these texture features