Efficient mode finding for retinal image segmentation via a new similarity measure

Abstract

Mean shift algorithm (MS) is an automatic method not only for finding mode of density for a given data but also for image segmentation. However, its weakness is that the MS is an expensive computation, especially in the retinal image which has very high data point. One obvious optimization is to avoid the redundancy computations. Hence, the main objective of this paper is to show that MS via a new similarity measure is most appropriate for retinal image segmentation. A new similarity evaluation of the retinal image was used for segmentation and then compared its results with the results from standard MS. Our results indicate that MS via a new similarity measure is a successful method in retinal image segmentation, reducing redundancy computation, and achieving speed up for 23.65 times faster than the standard MS.