A sea-land segmentation scheme based on statistical model of sea

Efficient sea–land segmentation using seeds learning and edge directed graph cut

“…• SMS: Global thresholding method (You and Li, 2011). It builds a statistical model for the sea based on the Otsu results.…”

“…When building prior models, Gaussian mixture model (GMM) can be employed as a proper description of the distribution of the sea (You and Li, 2011;Rother et al, 2004). We don't directly use the probabilistic outputs of SVM as the likelihoods of superpixels, because for an image, there is misclassification on the superpixels from SVM.…”

“…The method is suitable for images with coarse spatial resolution, while the convergence of the fitting process to a global optimum is not guaranteed. You and Li (2011) computed the maximum area of the sea based on Otsu results and built a statistical model for the sea area to decide the segmentation threshold. Zhu et al (2010) enhanced the original image with Sobel operator and segmented the enhanced image by Otsu method.…”

Semi-Automated Shoreline Extraction in Satellite Imagery and Usage of Fractals as Performance Evaluator

“…These studies uses supervised machine learning to classify land and sea by using remote sensing image features [5]. There are statistical models to perform segmentation [6]. Li and Yu [6] use local boundary optimization to extract shoreline.…”

“…There are statistical models to perform segmentation [6]. Li and Yu [6] use local boundary optimization to extract shoreline. This method uses OTSU [7] segmentation to perform a rough estimation of the statistical properties of the land and sea and a better segmentation is performed by using this assumption and estimation.…”