Region merging using local spectral angle thresholds: A more accurate method for hybrid segmentation of remote sensing images

详细信息    查看全文

• 作者：Jian Yang^a ; ^b ; ^{jiangeo.yang@mail.utoronto.ca} ; Yuhong He^b ; John Caspersen^c
• 关键词：Image segmentation ; Region merging ; Local SA thresholds ; Primitive segments ; Multi-band watershed transformation ; Geographic object-based image analysis (GEOBIA)
• 刊名：Remote Sensing of Environment
• 出版年：2017
• 出版时间：1 March 2017
• 年：2017
• 卷：190
• 期：Complete
• 页码：137-148
• 全文大小：3838 K
• 卷排序：190

文摘

We propose a hybrid segmentation algorithm by using edge- and region-based methods. Initial segmentation is produced by multi-spectral watershed transformation. The iteration of region merging is controlled by local spectral angle thresholds. We test the method performance by segmenting three images from different sensors. The success in three segmentation applications ensures the method transferability.