A Narrow River Extraction Method Based on Linear Feature Enhancement in TM Image

Extraction of narrow rivers from TM images is challenging due to mixed-pixel effects. This paper presents an automatic approach for narrow river extraction by integrating multiple digital image processing techniques. Firstly, the threshold segmentation was applied on water-index images to separate from planar water bodies and narrow rivers. Secondly, a linear feature enhancement algorithm is adopted to highlight river information and suppress other information. Thirdly, narrow rivers are extracted using dual-threshold line tracking method. Finally, three methods are selected to remove shadow, roads and other noises. Experimental results show the approach can effectively extract narrow rivers with the producer's accuracy higher than 82%, user's accuracy higher than 93%, Kappa coefficient higher than 0.993, and completeness higher than 90%, and avoid impact from multiple kinds of noise.