Multiresolution image denoising for detection of bridges in Satellite Images

Satellite Images contain very important information about geographical features such as roads, rivers, bridges, buildings, etc of earth. Such features are required for military as well as civilian applications. Automatic detection of these features are required for terrain mapping and Geographic Information System (GIS). However, satellite images are often corrupted by different types of noise and detection of details is a nontrivial problem. Hence suitable denoising algorithms have to be incorporated for the images before detection of features is done. Otherwise the process may end in erroneous results and loss of wanted information. Although different algorithms are existing for denoising images, due to the requirement of detection of long edges in the case of bridge extraction, the Curvelet Transform (CT)based method is found to be superior. This choice provides the benefit of multiresolution image processing such as compatibility of human visual system and line and edge detection. Once this pre processing is done, appropriate types of segmentation algorithms can be applied for detection process. Initially, Fuzzy based algorithm is used for segmenting water and background details and then bridge extraction is done using difference in pixel intensity between bridge edge and water.