Automatic Coastline Detection in Non-locally Filtered TanDEM-X Data

Abstract

The detection of coastlines in SAR imagery has been studied for more than two decades now. Whereas the first works were based on the exploitation of amplitude imagery and the corresponding need to deal with speckle noise [1, 2, 3], with the ERS-1/2 tandem configuration also coherence maps started to be used as input [4, 5]. Based on the insights gained on these experiments, later the authors began to exploit both amplitude and coherence imagery simultaneously [6, 7], finally giving way to the first approach using the original complex SAR data for statistically motivated coastline extraction [8]. In any case, one of the core problems for coastline detection from SAR datasets is the comparably high level of noise, which degrades the grayscale input imagery, be it amplitude or coherence. In order to solve this deficiency, an abundance of filtering techniques have been proposed over the years, among which the most sophisticated are based on the recently introduced non-local paradigm. Since Deledalle et al. proposed the NL-InSAR algorithm designed for interferometric multilooking in 2011 [9], novel denoising qualities can be achieved. In [10] and [11], it was already shown that the final DEM to be created during the German TanDEM-X mission [12] can be improved significantly by a specifically optimized implementation of NL-InSAR. This paper proposes a fully automatic framework for coastline extraction from non-locally filtered TanDEM-X data based on unsupervised active contours. It is therefore structured into two main parts: First, the non-local filtering framework will be described. Second, the active contours algorithm for coastline detection will be described. Finally, some preliminary experimental results processed from an interferometric TerraSAR-X dataset showing the skerry coast in the area of Stockholm, Sweden, will be demonstrated.