Automated Segmentation of Low Contrast Endothelial Cell Images for Fluorescent Intensity Measurement

An automated segmentation method is proposed for low contrast endothelial cell images. Morphological reconstruction and histogram equalization are used to remove the uneven background and enhance the contrast. Conventional thresholding and morphological filtering follow up to perform the coarse segmentation. Finally the watershed transform with markers obtained from the coarse segmentation is adopted to complete the segmentation. Average fluorescent intensity over the segmented cell regions is subsequently computed. Experimental results show that the proposed segmentation approach promotes the accuracy of the fluorescent intensity measurement significantly.