Using Morlet Wavelet for Retinal Vessel Segmentation

Extracting retinal vessels from the retinal image is important to the diagnosis of diabetic retinopathy and glaucoma in clinic. In terms of special gray distribution and regional configuration of retinal blood vessel network, this paper proposed a new blood vessel tree segmentation method based on Morlet wavelet and Gaussian matched filter. To construct the feature vector image, the Morlet wavelet cofficient responses are computed for the retinal vessel. Length filtering and hysteresis thresholds are subsequently implemented for the vessel tree extraction. The experimental results show that the proposed unsupervised method can effectively extract the blood vessels network on fundus image. Compared with other methods, our method has stronger robustness with less vessel stickiness, which is more practically valuable for the clinic diagnosis.