Effects of Color Constancy Algorithm on Retinal Blood Vessel Segmentation

One of the important steps towards a fully automated retinal image diagnosis system is automatic segmentation of blood vessels from the retinal image. In this paper, we investigate the effects of performing color constancy algorithm on retinal image prior to segmentation step. We compared two different algorithms for color constancy, namely Gray World and White Patch that are applied to retinal image before it is used as an input to the segmentation algorithm. We calculated the performance metrics for the outputs that includes Accuracy, Sensitivity, Specificity and MCC values to compare the performance with and without color constancy applied. From the results, it can be seen that color constancy improves the Sensitivity and MCC value of the segmentation output, with a small decrease in Specificity and Accuracy. Qualitatively, the outputs with color constancy applied have more small vessels, which were mostly left undetected in outputs with no color constancy. These small vessels are one of the key indicators for early detection of diabetic retinopathy.