Abstract:

Reliable and accurate retinal vessel segmentation and extraction is an important step in computer aided diagnosis systems in retinopathy. We have developed a supervised random forest (RF) classifier for segmenting thin vessel structures using multiscale features bank which are optimized for obtaining good segmentations on detecting faint vessel structures that would be missed traditionally with standard techniques. Experimental results on publicly available datasets DRIVE, STARE, CHASE, IOSTAR, HRF, VAMPIRE show that the our approach produces better results than some of the state of the art vessel segmentation methods when compared to manual ground truth gold standards.