Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
8145582 | Infrared Physics & Technology | 2018 | 18 Pages |
Abstract
Sea-land segmentation is a key step for the information processing of ocean remote sensing images. Traditional sea-land segmentation algorithms ignore the local similarity prior of sea and land, and thus fail in complex scenarios. In this paper, we propose a new sea-land segmentation method for infrared remote sensing images to tackle the problem based on superpixels and multi-scale features. Considering the connectivity and local similarity of sea or land, we interpret the sea-land segmentation task in view of superpixels rather than pixels, where similar pixels are clustered and the local similarity are explored. Moreover, the multi-scale features are elaborately designed, comprising of gray histogram and multi-scale total variation. Experimental results on infrared bands of Landsat-8 satellite images demonstrate that the proposed method can obtain more accurate and more robust sea-land segmentation results than the traditional algorithms.
Keywords
Related Topics
Physical Sciences and Engineering
Physics and Astronomy
Atomic and Molecular Physics, and Optics
Authors
Sen Lei, Zhengxia Zou, Dunge Liu, Zhenghuan Xia, Zhenwei Shi,