A wavelet transform method to detect boundaries between land and water in SAR image

Edge detection is important and key to image segmentation, information extraction, mapping etc. It is more difficult with SAR images than optical images, because of speckle. The aim of the paper is to show how boundaries between land and water can be detected from SAR images by using wavelet transform method, block-tracing algorithm and snake algorithm. The Ku-Band SAR image was obtained on 14 July 2003. The test area is near Huai River, located in city of Fuyang, Anhui Province. This SAR is made by Institute of Electronics, Chinese Academy of Sciences. Its azimuth resolution is 1.25 m and its range resolution is 1.2 m. To detect water edge, we apply a wavelet transform method to obtain the low-resolution image of the original image and suppress the speckle. Then we use a block-tracing algorithm to determine the coarse boundary between land and water and obtain a continuous edge. In the end, we use the snake algorithm to obtain the accurate edge. The result proves our method is a good method