A comparative analysis of watershed and edge based segmentation of red blood cells

Abstract

Background: Useful information which is helpful in the diagnosis of various disorders is obtained from the analysis of individual blood cells. Aim: To perform a comparative analysis between edge-based segmentation and watershed segmentation on images of the red blood cells. Method: The images to be used for the analysis were gotten from published research works. A database that contains both images was created in the Matlab environment. Edge-based segmentation and watershed segmentation were performed on the images. The edge based segmentation involves finding ridges, lines and contours along the images, while the watershed segmentation involves opening and closing reconstruction of overlapping features in images. MATLAB 2010a was used as the tool box for performing the segmentation process. Result: The value of correlation for segmentation with edge and watershed stood at 0.9499, 0.9198, respectively. Also, in terms of deviation, values for both were 65.846, and 60.317 respectively, and also in terms of area, the values were 39520 and 4467 respectively. Finally, in terms of mean, the values were 187.06 and 17.006. Conclusion: Watershed segmentation outperforms edge based segmentation in terms of image statistics and performance, which can help physician and medical practitioners to identify possible blood disorder.

Key words: Watershed, red blood cell, segmentation, edge-based, Matlab, reconstruction