Coastline detection in SAR images using texture analysis in textural or geometrical multi-resolution

A new method is proposed in this paper for coastline extraction which uses the SAR amplitude images, or if available, the interferometric coherence images. This method uses texture analysis based on textural multi-resolution (TMR) or geometrical multi-resolution (GMR), for a fine analysis of coastline. First order texture measures (TM) have been used. In TMR, a series of TM are generated at various textural scales from the original image. In GMR case, a wavelet transform produces a low-resolution approximation of the original image. Consequently, TM are also generated at a suitable low-resolution approximation. A selection of the suitable TM in one or another multi-resolution approach, leads to a good separation of sea and coastal area and reveals coastline. A detection of discontinuities allows then easy extraction of coastline. The obtained results are promising and show a good detection, and a good continuity of coastline. A proposed method has been tested on ERS SAR amplitude and coherence coastal region images of Cameroon, Mauritania and Guyana.