synthetic Retinal Image Sequence Registration (RISeR-syn) dataset

RISeR-syn is generated using the method described in our paper. We select 10 retinal images which are from different eyes, and take one of them as a given reference image each time to generate 5 different types of synthetic retinal image sequences, resulting in a total of 50 sequences, each consisting of 6 synthetic retinal images. Compared with RISeR-real, which is time-consuming and prone to introducing errors when manually annotating control points, RISeR-syn can easily obtain densely distributed and accurately corresponding control points when it is generated, thereby achieving a more objective and effective registration evaluation. Moreover, under the premise of taking the first image in sequence as the reference frame, we also annotate control points that only exist in the common retinal area between two adjacent frames but not in the reference frame, in order to more comprehensively evaluate the inter-frame continuity of the sequence registration results. Note: A more detailed description can be found in our upcoming paper, coming soon!