Automatic road extraction using the Mean Shift approach

Analysis of remotely sensed images for road extraction is becoming a vital feature for urban planning and applications based on visual data. It is also one of the key concept of intelligent automobile guidance. There is a need for fast road extraction methods using satellite remote sensing techniques. The Mean Shift approach for the road extraction is exploited herein. The input satellite image is preprocessed to increase the tolerance of the image and to remove noise due to non-road structures. The median filtering technique is used to extract the linear segments of the road (road structures are basically a group of many smaller linear structures). The resultant images after the road extraction operation are evaluated based on quality measures and then quantitatively compared with the results of previous approaches