Ferrographic image segmentation by the method combining k-means clustering and watershed algorithm

By evaluating the results of k-means clustering in different color spaces including RGB, HSI and CIELAB, this study proposed the algorithm of k-means clustering using two dimensional color components in CIELAB color space. By this algorithm, the wear particles could be segmented directly from the background of ferrographic image. Then, the results of k-means clustering are used as basic images, threshold method is adopted to extract regional minimal values of particles and background to obtain the marker images of both particles and background. At last, the automatic segmentation of wear particles is achieved by using improved watershed algorithm. The results show that the method in this study could improve the segmentation accuracy of wear particle chains by eliminating the influences from background.