Phase-shift and information measures for retinal motion detection and compensation with optimized parameter selection

Infrared images sequences of the retina are commonly used for diagnosing diabetic macular edema(DME) and its treatment. Retinal motion detection and its compensation are important steps in the laser treatment of DME. In this paper, we propose to use the phase shift approach and information measures for detecting and compensating retinal motion in infrared image sequences, despite the fact that parameter selection and tuning is a drawback of this approach. Phase Shift and related information measures usually requires several parameter adjustments, and we show, based on the analysis of variance, that the number of parameters needing fine tuning can be reduced and still have reliable retinal motion detection and its compensation in noisy infrared image sequences. Our preliminary experimental results indicate that high positive predictive values can be achieved with a reduced number of critical parameters.