A new retinal vessel segmentation method using preprocessed Gabor and local binary patterns

Abstract

A new retinal vascular tissue segmentation algorithm, which utilizes Gabor wavelet and local binary patterns, is introduced. It would be shown that how a simple preprocessing step would increase the accuracy of algorithm. Different features have been proposed for retinal vessel detection. One of the most famous features adapted is Gabor wavelet. Thanks to multi-resolution property of Gabor, combination of scales can be used to extract features. However, similar features in feature vector would increase the inter-correlation and may lead to poor result. Also, Local Binary Pattern (LBP) is applied. LBP is a powerful feature for texture analysis. A wise pre-processing strategy is applied to image with regard to feature extraction technique. Contrary to previous methods where a simple pre-processing scheme applied for all feature extraction methods, here each feature extraction will utilize its own suitable preprocessing. It is showed that this enhances the result of segmentation. The proposed method has a low dimension feature vector having only four features. The pre-processing step enhances the results in comparison to a previous method in term of area under the ROC curve The computational results of simulations show the high performance of the proposed method in term of accuracy and speed.