Human retina optic disc segmentation using statistical region merging

Optic disc (OD) localization and segmentation are important tasks in automatic eye disease screening. In this thesis we presented a new, fast and simple iterative methodology for semi-automatic localization and segmentation of the optic disc in fundus images. Furthermore, this new method can find the area of optic disc using the statistical region merging algorithm. The proposed method uses Matlab programming languages for evaluation of algorithm. The performance of the proposed method compared with various methods in the literature, and the results are found convincing and efficient. The obtained results indicate that this method of the segmentation of OD has good accuracy. This thesis presents information about human retina optic disc and shows the most common problems that may happen on it. Moreover, in this thesis discuss deal with the symptoms of the diseases that may affect human eyes and how to detect these diseases. In addition, in this thesis we gived the previous techniques in image segmentation which were used before to capture and display the image for both diagnose and therapeutic purpose. In other words, we explained what the image processing, biomedical image processing, and the steps of image processing. Then we will discuss the most important step in image processing which is image segmentation, and the techniques have already been used before in image segmentation. Finally, a new technique or new algorithm has been proposed to address the disadvantages of the prior art and to improve the performance and efficiency of image segmentation on the Human Retina Optical Disc.In contrast to the classical techniques, this new method has more effective results with more insight, fast, high precision and more effective time. Initially, OD position candidates were identified using the median filter and the Otsu method. After locating the optical disc, statistical region joining method was applied to this region. For the algorithm and calculations, MatLab version 16 was used. All retinal images in this study were taken from the public and international MESSDIOR database.