This paper presents a method for automated segmentation of retinal vessels using the 2-D Morlet wavelet and supervised classification. The method classifies each image pixel as vessel or non-vessel based on a feature vector composed of pixel intensity and continuous 2-D Morlet wavelet transform responses at multiple scales. The Morlet wavelet is capable of tuning to specific frequencies, allowing noise filtering and vessel enhancement in a single step. A Bayesian classifier with Gaussian mixture models is used, which provides fast classification and can model complex decision surfaces. The method is evaluated on the DRIVE and STARE databases, achieving an area under the ROC curve of 0.9598 on the DRIVE database, slightly better than the method of Staal et al. The method involves pre-processing to remove border effects, applying the 2-D Morlet wavelet transform, and using supervised classification to distinguish vessel and non-vessel pixels. The feature space is normalized to ensure consistent classification. Two classifiers are tested: a Gaussian mixture model (GMM) classifier and a linear minimum squared error (LMSE) classifier. The GMM classifier provides better performance but requires more training time, while the LMSE classifier is faster but less accurate. The method is evaluated using ROC curves, which measure true positive and false positive fractions. The GMM classifier achieves an area under the ROC curve of 0.9598 on the DRIVE database and 0.9651 on the STARE database. The results show that the method performs well compared to manual segmentations, with the GMM classifier outperforming the LMSE classifier in accuracy and performance. The method is effective in enhancing vessel contrast while filtering out noise. It is capable of segmenting vessels of different diameters using Morlet transforms at different scales. The method is trained using manual segmentations and can be applied to different types of images. The results suggest that the method is a promising approach for automated retinal vessel segmentation, with potential for further improvements through post-processing and the use of smaller training sets. The method is also suitable for semi-automated segmentation, where only a small portion of the image needs to be manually segmented. The results indicate that the method is effective in detecting vessels, although there are some challenges in detecting noise and other artifacts. The method is also capable of analyzing the vessel pattern of the ocular fundus using nonlinear methods such as fractal analysis. The method is a promising approach for automated retinal vessel segmentation, with potential for further improvements through post-processing and the use of smaller training sets.This paper presents a method for automated segmentation of retinal vessels using the 2-D Morlet wavelet and supervised classification. The method classifies each image pixel as vessel or non-vessel based on a feature vector composed of pixel intensity and continuous 2-D Morlet wavelet transform responses at multiple scales. The Morlet wavelet is capable of tuning to specific frequencies, allowing noise filtering and vessel enhancement in a single step. A Bayesian classifier with Gaussian mixture models is used, which provides fast classification and can model complex decision surfaces. The method is evaluated on the DRIVE and STARE databases, achieving an area under the ROC curve of 0.9598 on the DRIVE database, slightly better than the method of Staal et al. The method involves pre-processing to remove border effects, applying the 2-D Morlet wavelet transform, and using supervised classification to distinguish vessel and non-vessel pixels. The feature space is normalized to ensure consistent classification. Two classifiers are tested: a Gaussian mixture model (GMM) classifier and a linear minimum squared error (LMSE) classifier. The GMM classifier provides better performance but requires more training time, while the LMSE classifier is faster but less accurate. The method is evaluated using ROC curves, which measure true positive and false positive fractions. The GMM classifier achieves an area under the ROC curve of 0.9598 on the DRIVE database and 0.9651 on the STARE database. The results show that the method performs well compared to manual segmentations, with the GMM classifier outperforming the LMSE classifier in accuracy and performance. The method is effective in enhancing vessel contrast while filtering out noise. It is capable of segmenting vessels of different diameters using Morlet transforms at different scales. The method is trained using manual segmentations and can be applied to different types of images. The results suggest that the method is a promising approach for automated retinal vessel segmentation, with potential for further improvements through post-processing and the use of smaller training sets. The method is also suitable for semi-automated segmentation, where only a small portion of the image needs to be manually segmented. The results indicate that the method is effective in detecting vessels, although there are some challenges in detecting noise and other artifacts. The method is also capable of analyzing the vessel pattern of the ocular fundus using nonlinear methods such as fractal analysis. The method is a promising approach for automated retinal vessel segmentation, with potential for further improvements through post-processing and the use of smaller training sets.