Optic Disc Localization Based on Feature Sorting
Localization of the optic disc (OD) is a necessary step in automatic diagnosis of ocular diseases in retinal images: diabetic retinopathy, glaucoma and so on. In this paper, we combine different features and classification schemes to increase the performance of OD detection and localization. To this...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Publishing House of the Romanian Academy
2016-09-01
|
| Series: | Memoirs of the Scientific Sections of the Romanian Academy |
| Subjects: | |
| Online Access: | http://mss.academiaromana-is.ro/mem_sc_st_2016/6_Ichim.pdf |
| Summary: | Localization of the optic disc (OD) is a necessary step in automatic diagnosis of ocular diseases in retinal images: diabetic retinopathy, glaucoma and so on. In this paper, we combine different features and classification schemes to increase the performance of OD detection and localization. To this end, we propose a simple image processing algorithm based on adaptive local texture analysis considering different features, such as those extracted from the co-occurrence matrix, the fractal dimension and blood density. The selection of features is made in the learning phase, taking into account their relevance and non-redundancy. Retina images are decomposed in patches using the sliding box method. The presence of regions with different intensities and noise requires preprocessing operations. For OD recognition, a method which combines a voting scheme with a sorting procedure is applied. In the experiments, 100 images from the publicly available STARE dataset were used. |
|---|---|
| ISSN: | 1224-1407 2343-7049 |