The aim of this study is to help ophthalmologists and opticians during the diagnostic process of the retinal fundus. We propose a computer-vision-based solution that allows, from stereo images, the extraction of clinical parameters and/or the generation of multi-viewpoint images of the retinal fundus. This goal can be achieved by estimating the disparity map of the stereo images. For more precise clinical parameter extraction, a sub-pixel approach could be used. Additionally, the a priori knowledge of the fundus geometric shape provides useful information for the disparity map estimation process. In this paper we propose a sub-pixel disparity estimation algorithms that takes into consideration the geometric shape of the retinal fundus. Different stereo images, with known and unknown ground truth, are used to compare the proposed algorithms to state-of-the-art algorithms and to demonstrate the efficiency of our proposed method. © 2016, Society for Imaging Science and Technology (IS&T).