An adaptive thresholding approach for automatic optic disk segmentation
This work addresses a key step in automatic retinal screening for medical diagnosis, but it appears incremental as it builds on existing thresholding methods.
The paper tackles optic disk segmentation in retinal images by proposing an adaptive thresholding algorithm that handles illumination variations, achieving correct location in all images on the DRIVE database with 43.21% mean overlap and 98% detection on KHATAM with 36.32% mean overlap.
Optic disk segmentation is a prerequisite step in automatic retinal screening systems. In this paper, we propose an algorithm for optic disk segmentation based on a local adaptive thresholding method. Location of the optic disk is validated by intensity and average vessel width of retinal images. Then an adaptive thresholding is applied on the temporal and nasal part of the optic disc separately. Adaptive thresholding, makes our algorithm robust to illumination variations and various image acquisition conditions. Moreover, experimental results on the DRIVE and KHATAM databases show promising results compared to the recent literature. In the DRIVE database, the optic disk in all images is correctly located and the mean overlap reached to 43.21%. The optic disk is correctly detected in 98% of the images with the mean overlap of 36.32% in the KHATAM database.