부산대학교 도서관

We present a system of automated drusen detection from color fundus photographs with our ultimate goal being to automatically assess the risk for the development of Age-related Macular Degeneration (AMD). Our system incorporates learning based drusen detection and includes fundus image analysis techniques for image denoising, illumination correction and color transfer. In contrast to previous work, we incorporate both optimal color descriptors and robust multiscale local image descriptors in our drusen detection process. Our system was evaluated with color fundus photographs from two AMD clinical studies [1, 2]. By comparing our results to those obtained via manual drusen segmentation, we show that our system outperforms two state-of-the-art techniques.