The technique developed in this project maps quality criteria inspired by diagnosis procedures based on the advise of an eye expert to quantitative and objective features related to image quality. Independent from segmentation methods it combines global clustering with local sharpness and texture features for classification. The clustering describes the recognizability, dissimilarity and contrast of relevant structures. The sharpness metric evaluates the separation between components. The texture features measure common image sharpness, homogeneity and generic contrast.

On a test dataset of 301 retinal fundus images the method was evaluated on a given gold standard by human observers and compared to a state of the art approach. An area under the ROC curve of 95.3% compared to 87.2% for correctly identifying bad quality images outperformed the state of the art approach. A significant p-value of 0.019 emphasizes the statistical difference of both approaches.

AQuAFo (GUI for Automated Quality Assessment of Retinal Fundus Photos) provides a Graphical User Interface (GUI) implementation of this algorithm for free usage and is available here.