A novel method for vessel detection using Contourlet Transform

Vessel detection is an important task for diagnosis of vascular diseases in clinical images. Many diseases such as diabetic retinopathy and hypertension can be detected by retinal vessel map or scanning conjunctival vessels. There are a lot of techniques for vessel extraction from retinal images but most of them have failed to face with some patterns like hemorrhages and micro aneurysms. In this paper we develop an algorithm based on Non-subsampled Contourlet Transform (NSCT) and morphological operations. By combining of information from two scales of contourlet and gray scale image, vessel map is extracted. Optic disc border is eliminated by Non-subsampled Contourlet directional information. In addition, circular shapes such as micro aneurysms are removed using morphological operations. We examine our algorithm on retinal images of DRIVE database and conjuntival images of Khatam database. Experimental results show significant improvements in achieving high accuracy and decreasing False Positive Rate (FPR) of vessel detection on both databases.