A Retinal Image Idetification Method using Blood Vessel Topological Information


Medical Technologies National Congress (TIPTEKNO), Trabzon, Turkey, 12 - 14 October 2017 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/tiptekno.2017.8238121
  • City: Trabzon
  • Country: Turkey
  • Karadeniz Technical University Affiliated: Yes


In recent days biometric features are frequently used in the field of personal identification and authentication. Retinal vessels are important biometric features in this manner. It is suggested a personal identification system used retinal blood vessel topology in this paper. For this purpose, characteristic feature matrix that is extracted by using topological features. The position of these feature points is found by using vessel skeleton structures and the adjacency matrix is extracted. Then a characteristic feature matrix that represent the retinal image is obtained. Sixty binary retinal images obtained from STARE and DRIVE databases available on the internet are used and sixty affine transformed images are newly obtained. Totally 120 retinal images are cross-mixtured and 60 test images are searched one by one in the 60 other images. As a result, about 95% identification success rate is achieved.