A NOVEL SEGMENTATION METHOD BASED ON GRAYSCALE WAVE FOR UNDERWATER IMAGES

摘要

The segmentation of underwater images is quite difficult due to the effects by water medium and non-uniform illumination. A local threshold segmentation method based on image grayscale waves is proposed in this paper. First, an unsymmetrical median filter (UMF) is applied to reduce noise and improve edge features. Second, large scale troughs and crests on the grayscale wave along row and column directions are detected for locating the edges and the bodies of foreground objects in the image. Finally, the foreground objects are segmented from the background with local thresholds based on the large scale crests and troughs. Experimental analysis is conducted by using underwater images of different objects taken by an underwater vehicle with artificial illumination. The results show that the proposed method is capable of segmenting images of different contrast and uneven brightness.