A NEW IMAGERY CLASSIFICATION METHOD USING SPATIAL COVARIANCE INFORMATION: AN APPLICATION TO THE SAR IMAGE OF THE COASTLINE OF CAMEROON

摘要

Classical and modern statistical methods offer wide variety approaches of data classification in general and image classification in particular. Among these approaches none of them explicitly use spatial information. Structures with spatial covariance have been used for predicting data, but are not directly use for classifying. This paper deals with a supervised classification method based on spatial covariance information obtained by image texture analysis. An experimental variogram is plotted for each training zone, and fitted with empirical variogram models. The parameters deduced from these models are stored in a feature vector of texture. This method has been applied for the classification of a SAR image of the Atlantic coast of Cameroon.