Abstract: A new scheme to extract roads in urban area is presented by taking the advantages of several existing methods.The original IKONOS image is segmented using the angular texture signature principle.Then the line segment matching approach is utilized to remove non-road targets from the segmented result，so that roads boundaries appear more regular.In the third step central lines of all roads are obtained with the aid of some morphological algorithms，furthermore a curve central line is broken into piecewise lines.Finally central road line segments are grouped to restore the whole road network using Markovian model and context knowledge.Experimental results show that the proposed method is efficient for extracting roads in urban area from high-resolution images.