Texture analysis of terrain images using a Markov random field model

An approach to determining the textural attributes of an image an a pixel-by-pixel basis is presented. The technique is based on modelling the spatial interactions within the image using a Gaussian Markov random field. The method is applied to the classification of homogeneous texture images, artificial texture collages and a composite-texture SAR terrain image. The capabilities and limitations of the technique are considered in the light of these classification experiments.<
>