High-resolution Coastline Extraction in SAR Images via MISP-GGD Superpixel Segmentation

Odysseas Pappas,Nantheera Anantrasirichai,Byron Adams,Alin Achim

DOI: https://doi.org/10.48550/arXiv.2203.02708

2022-03-05

Abstract:High accuracy coastline/shoreline extraction from SAR imagery is a crucial step in a number of maritime and coastal monitoring applications. We present a method based on image segmentation using the Generalised Gamma Mixture Model superpixel algorithm (MISP-GGD). MISP-GGD produces superpixels adhering with great accuracy to object edges in the image, such as the coastline. Unsupervised clustering of the generated superpixels according to textural and radiometric features allows for generation of a land/water mask from which a highly accurate coastline can be extracted. We present results of our proposed method on a number of SAR images of varying characteristics.

Image and Video Processing,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

This paper aims to solve the problem of high - precision extraction of coastlines from Synthetic Aperture Radar (SAR) images. Specifically, the authors propose an image segmentation method based on the Meta - superpixel algorithm with Generalized Gamma Distribution (MISP - GGD), combined with unsupervised clustering techniques, to classify the generated superpixels according to texture and radiometric characteristics, thereby generating a land / water mask and finally extracting a high - precision coastline. This method is particularly suitable for marine and coastal monitoring applications that require accurate coastline information, such as mapping, automatic ship navigation, offshore vessel detection, and coastal erosion monitoring.

The paper mentions that due to the inherent speckle noise in SAR images and the low land - sea contrast in bad weather conditions, the coastline detection task becomes very challenging. Traditional coastline detection algorithms usually involve segmenting the image into two major categories of water and land, and then extracting a continuous single - pixel boundary through morphological processing. The MISP - GGD algorithm proposed in this paper can generate superpixels that highly match the edges of objects, which helps to more accurately depict complex linear structures, thereby simplifying the subsequent classification / clustering steps. In addition, this method does not require training data and can be executed on a single image without the need for interferometric image pairs or multi - polarization products. It has high robustness to data variability (such as polarization, incidence angle, and sea state), and does not involve filtering operations that may cause coastline blurring.