STATISTICAL TEXTURE ANALYSIS OF ROAD FOR MOVING OBJECTIVES

The paper objective is to identify and localize significant regions in textured images for navigating an autonomous agent. Two types of statistic texture features are used. The first type features are the histograms, and the edge density. The second type features derive from the medium co-occurrence matrices, which is a notion introduced by the authors. The algorithms are implemented in MATLAB. The basic texture is considered the asphalt and the different textures are considered the grass and the pebble. The experimental results indicate that the features, which derive from medium co-occurrence matrices, have a good discriminating power both for texture classification and region localization.