Research Of Retinal Blood Vessels Segmentation

The retinal vasculature is the only part of the body’s circulatory system that can be observed non-invasively by optical means,which its structural changes such as vessel diameter,bending degree and so on have a huge contact with many kinds of ophthalmic diseases.Besides,the uniqueness and imperceptibility and Collectability of retinal images can be also applied to the distinction between individual characteristics and personal identification.This makes the extraction of retinal vasculature a rewarding and well researched subject on clinical medicine and personal identification,security and secrecy.At present,there are many methods of the study of retinal blood vessels segmentation.On the basis of reading a large number of literatures of digital image processing and image segmentation,the article firstly has summarized and analyzed the existing schemes,and design two new retinal vessel segmentation methods based on the multiscales geometric transform and tensor voting,which focus on the noise and small and fractured vessels segmentation based on the structure feature of retinal vessels.(1)The method based on multiscales geometric transform:this thesis proposes a retinal blood vessel segmentation method based on multiscales geometric transform and OTSU threshold.Firstly,the green channel of image has been enhanced by the improving Curvelet filter which is adaptive.Then the improving OTSU threshold is applied to segment the vessel to get the extraction.(2)The method based on tensor voting:this thesis proposes a retinal blood vessel segmentation method based on tensor voting.Firstly,a preprocess has been used to get a binary image.Then the voting of tensor which has an information of orientation has been done,following the extraction of the salience.At last,a morphological processing has been applied to get the final vascular network.This thesis makes a lot of simulation experiments based on the REVIEW database and DRIVE database,and calculates the accuracy,sensitivity and 1-specificity of these two methods proposed by this thesis,compares with the results of manual and existing methods.Experiment results show that,this method plays a better performance on the accuracy,sensitivity and 1-specificity than existed methods.Moreover,methods proposed by this thesis can segment more small vessels,especially for the small vessels at the low gray contrast.