点击次数：

所属单位：电子信息工程学院

发表刊物：2017 IEEE 2ND INTERNATIONAL CONFERENCE ON SIGNAL AND IMAGE PROCESSING (ICSIP)

关键字：optic disc localization deep learning convolution neural network saliency map

摘要：The optic disc (OD) contains lots of important information in retinal image analysis. Detecting the region of OD correctly is important for subsequent analysis of retinal images. It is challenging to locate the OD precisely due to the various reasons including low image quality or lesions around the OD. In this paper, we propose a cascading localization method based on deep learning with feedback to improve the accuracy of OD localization. The method employs the model of saliency-based visual attention to find the most salient region and implements deep convolution neural network (CNN) to determine whether it contains OD. If a region is classified as non-OD region, we find the next salient region and input it into the CNN to classify. The algorithm ends when the CNN finds a region with OD. The proposed method is evaluated on the ORIGA and MESSIDOR datasets. Our experimental results show that the proposed method achieved significant improvement in detection of OD compared with previous methods.

是否译文：否

发表时间：2017-01-01

合写作者：Niu, Di,Xu, Peiyuan,Cheng, Jun,Liu, Jiang

通讯作者：万程