OPTIC DISC SEGMENTATION IN RETINAL FUNDUS IMAGES
Author(s):
Nabi Amin1, Salman Khan2, Tanveer Hussain3, Siraj4, Zahoor Jan5, Muhammad Sajjad6
1 Postgraduate student, Department of Computer Science, Islamia College Peshawar, Pakistan, Ph. +923459179747, Email: icup.amin@gmail.com.
2 Undergrad student, Department of Computer Science, Islamia College Peshawar, Pakistan, Ph. +923139608016, Email: salmank255@gmail.com.
3 Undergrad student, Department of Computer Science, Islamia College Peshawar, Pakistan, Ph. +9232290927696, Email: tanveer@icp.edu.pk.
4 Postgraduate student, Department of Computer Science, Islamia College Peshawar, Pakistan, Ph. +923042904056, Email: siraj.scholar@icp.edu.pk.
5 Associate Professor, Department of Computer Science, Islamia College Peshawar, Pakistan, Ph. +923339051566, Email: Muhammad.sajjad@icp.edu.pk.
6 Assistant Professor, Department of Computer Science, Islamia College Peshawar, Pakistan, Ph. +923339319519, Email: zahoor.jan@icp.edu.pk.
Volume:
Thematic Issue on Advances in Image and Video Processing
Pages:
25 - 37
Date:
May 2018
Abstract:
This paper presents efficient optic disc segmentation algorithm which integrates the optical disc (OD) pixels information from retinal fundus images. The method is based on the homomorphic system along with the automatic seed point region growing technique which provided good results despite of variance in shape, size and uneven illumination of different resolution retinal images. The method consists of three stages. The images are are de-noised by mathematical morphology (median filter) in the first stage. The homomorphic system is applied to increase and decrease the contrast of high and low frequencies, respectively, using the high-pass filter. Finally, the centre of OD detection is carried out by the Otsu’s thresholding technique which is used as input for region growing algorithm. Preliminary results showed good accuracy of the proposed system as compared to other existing methods.