Automatic K-Means for Color Enteromorpha Image Segmentation
Liang Qu, Xinghui Dong, Fadong Guo · 2009
In this paper, we introduce a simple automatic color enteromorpha image segmentation algorithm. First, the color images are converted from RGB into NTSC color space. Then, we scale the data of the saturation channel in NTSC color space to the range of 0-255 and obtain its histogram. Using this histogram, we can obtain two peaks in the enteromorpha and background class respectively. Thus, two positions in these two classes can be obtained. Thirdly, those two positions are used as the centroids in the k-means algorithm. By means of k-means algorithm, every enteromorpha image can be divided into two classes: enteromorpha and background class. In fact, it is only a pre-processing for enteromorpha detection. Experimental results show that our approach can segment the enteromorpha images very accurately.