Image Metrics for Predicting Subjective Image Quality


    loading  Checking for direct PDF access through Ovid

Abstract

Purpose.Despite the proliferation of wavefront sensors to characterize the optical quality of individual eyes, there is not yet an accurate way to determine from a wave aberration how severely it will impact the patient's vision. Some of the most commonly used metrics, such as RMS wavefront error and the Strehl ratio, predict subjective image quality poorly. Our goal is to establish a better metric to predict subjective image quality from the wave aberration.Methods.We describe three kinds of experiments designed to compare the effectiveness of different metrics in determining the subjective impact of the wave aberration. Subjects viewed a visual stimulus through a deformable mirror in an adaptive optics system that compensated for the subject's wave aberration. In the first experiment, we show that some Zernike modes such as spherical aberration and defocus interact strongly in determining subjective image quality. In the second experiment, the subject's wave aberration was replaced by the wave aberration corresponding to an individual Zernike mode. The subject then adjusted the coefficient of the Zernike mode to match the blur of a standard stimulus. In the third experiment, the subject viewed the same stimulus through the wave aberration of one of 59 different postoperative patients who had undergone LASIK and matched the blur by adjusting defocus. We then determined which among many image quality metrics best predicted these matching data.Results.RMS wavefront error was a poor predictor of the data, as was the Strehl ratio.Conclusions.The neural sharpness metric best described the subjective sharpness of images viewed through the wave aberrations of real eyes. This metric can provide a single number that describes the subjective impact of each patient's wave aberration and will also increase the accuracy of refraction estimates from wavefront-based autorefractors and phoropters.

    loading  Loading Related Articles