Extraction of Curvilinear Features in High-resolution Multispectral Satellite Images

Satellite imagery provides an effective aid to city planning. From a high-resolution spatial image, useful information such as road networks may be extracted readily. In this work we study a method to automate the extraction of curvilinear objects such as roads and canals from the very-high resolution multi-spectral IKONOS satellite imageries. Curvilinear feature extraction is composed of two steps: detection and connection. The two steps can be modular but they are not independent. The first step of work has been reported in (Hui, Liew and Kwoh 2001). In this paper, we present some of our exploration on the linking-up of the detected features with the idea of perceptual organization.