An effective approach towards color image segmentation for micro-vessel detection

In this paper, we propose an effective approach towards color image segmentation for micro-vessel detection in the virtual slide of double stained liver tissue. The contribution of the proposed method lies in three aspects. First, the dominant colors of white/gray, blue, red and brown are modeled in the RGB color space. Second, the CART (classification and regress tree) method is implemented to segment the modeled colors. Third, the micro-vessel regions are identified in the color images based on the association of both red and brown pixels. Extensive experiments are carried out to validate the performance of the proposed approach. The experimental results demonstrate that our method is quite promising.