Extraction of Roads from High-Resolution Satellite Images with the Discrete Wavelet Transform

Abstract

In this paper, a wavelet-based approach for road extraction from satellite images is proposed. The discrete wavelet transform (DWT) is used in the proposed approach to decompose an image into approximation and detail components in order to reduce the dimensions of unwanted small objects in the image due to the decimation effect of the DWT. After that, two morphological direction filters and a length filter are used with the approximation component. The objective of the two morphological filters is to detect roads in different directions and the objective of the length filter is to remove noise and small unwanted objects based on a certain threshold length. With the decimation effect of the DWT, it is expected that the length filter will succeed in removing most unwanted objects, easily. Images of high resolution satellites such as QuickBird with 0.61 m resolution, IKONOS with 1 m resolution, and SPOT-5 with 2.5 m resolution for different regions of Egypt and some other areas are used for the validation of the proposed approach.