Retinal Image Registration based on Auto-Adaptive SIFT and Redundant Keypoint Elimination Method
Zahra Hossein‐Nejad, Mehdi Nasri · 2019
Retinal image matching is a main step in many applications, including retinal image registration and retinal image mosaicking. The A2SIFT algorithm is one of the improved versions of SIFT algorithm, originally proposed for matching aerial imagery. One of the most important weaknesses of A2SIFT algorithm in retinal image matching is the contrast threshold value and the presence of a redundant feature. In this paper, first the contrast threshold value of this method has been modified and optimized for retinal image matching. After that, RKEM algorithm is used to increase its efficiency. The results of the experiments show the high efficiency and accuracy of the proposed method in retinal image matching.