Superpixel–Based SLIC and Single Threshold for Exudates Detection in Retinal Images
Abstract
The automated segmentation of exudates in color retinal images has been widely adopted for diagnosing ophthalmic. Exudate’s early screening can considerably reduce the patients’ loss of vision. The current paper proposes several highly efficient image processing methods to detect the exudates from color retinal images. One of the main challenges of this study is to deal with non‒preprocessing methods tasks while the removal of the optic disc is necessary for exudates detection because the optic disc pixels and a connected component could be considered as the exudates pixels. This novel candidate exudates detection is proposed by using a superpixel‒based simple linear iterative clustering (SLIC), a new contextual feature based on a single threshold, and optimal Otsu threshold. The proposed methods are evaluated on three different retinal image databases. The experimental results performed on poor quality retinal image reveal that the proposed the method offers excellent exudates detection outcomes using much less computational time, yet the results yield a high detection accuracy