We present REMPE (Registration of Retinal Images through Eye Modelling and Pose Estimation), a novel solution to the problem of retinal image registration. We solve it via simultaneously estimating the relative pose of the cameras that acquired the images as well as the shape and the pose of the eye. The method utilizes an ellipsoidal model for the eye, and the pose of the cameras is estimated utilizing RANSAC, followed by a variant of Particle Swarm Optimization (PSO). Extensive experiments demonstrate accurate and robust retinal image registration.

The full description of the registration method is explained in the following thesis:

Retinal Image Registration through 3D Eye Modelling and Pose Estimation
C. Hernandez-Matas
PhD thesis, University of Crete (Greece), 2017

If utilizing this executable, please cite:

REMPE: Registration of Retinal Images through Eye Modelling and Pose Estimation
C. Hernandez-Matas, X. Zabulis, A.A. Argyros
IEEE Journal of Biomedical and Health Informatics, 2020
DOI: 10.1109/JBHI.2020.2984483

Download REMPE-1.1.0 (8 MB)

The compressed file contains:

• readme.html: Readme file with information relative on how to configure and run the program.
• web: A folder containing support files for the readme.html file.
• REMPE.exe: The registration method executable.
• config.cfg: A sample config file.
• license.txt: The REMPE executable license file.
• libraries_README.txt: Relevant information regarding the libraries utilized.
• FIRE: A folder containing a sample image pair, as well as control points and masks of the FIRE dataset. The full dataset is publicly available here.

The program runs out of the box. Just decompress the file to a folder, double click on the executable and it will start registering the sample image pair provided, saving the results in the FIRE folder. A step-by-step guide is shown in the sanity test guide.

Detailed instructions can be found in the configuration instructions.

The program has CUDA GPU acceleration. It will automatically detect the properties of the installed graphic card (if any), and if compatible, it will run the program using GPU acceleration. Otherwise, CPU acceleration will be used. The program requires CUDA Compute Capability 3.0 or higher. A list with the Compute Capability of nVidia devices can be found in this link: CUDA GPUs.

If using a CUDA compatible card, please ensure that the drivers are updated to the latest version.

The Fundus Image Registration Dataset (FIRE), an ideal dataset to run experiments with this executable has been made publicly available at https://projects.ics.forth.gr/cvrl/fire/

The following mosaic images have been created utilizing the output provided by the program:

For any issue regarding this page or the executable, please contact: Carlos Hernandez-Matas https://carlos.hernandez.im

1. Retinal Image Registration through 3D Eye Modelling and Pose Estimation
   C. Hernandez-Matas
   PhD thesis, University of Crete (Greece), 2017
2. REMPE: Registration of Retinal Images through Eye Modelling and Pose Estimation
   C. Hernandez-Matas, X. Zabulis, A.A. Argyros
   IEEE Journal of Biomedical and Health Informatics, 2020
   DOI: 10.1109/JBHI.2020.2984483
3. Retinal image preprocessing, enhancement, and registration
   C. Hernandez-Matas, A.A. Argyros, X. Zabulis
   Computational Retinal Image Analysis: Tools, Applications and Perspectives, November 2019
   DOI: 10.1016/B978-0-08-102816-2.00004-6
4. FIRE: Fundus Image Registration Dataset
   C. Hernandez-Matas, X. Zabulis, A. Triantafyllou, P. Anyfanti, S. Douma, A.A. Argyros
   Journal for Modeling in Ophthalmology, vol. 1, no. 4, pp. 16-28, Jul. 2017.
   URL: https://www.modeling-ophthalmology.com/index.php/JMO/article/view/42

This research was made possible by a Marie Curie grant from the European Commission in the framework of the REVAMMAD ITN (Initial Training Research Network), Project 316990. It was also supported by the FORTH-ICS internal RTD Programme "Ambient Intelligence and Smart Environments".