Abstract
Recently, several research groups (Campbell CE, 2004; Thibos et al, 2004; Chen et al, 2005), including ours, have applied many different metrics of image quality for a variety of purposes. Campbell (Campbell CE, 2004) has previously defined three characteristics of a “good” metric: monotonicity, linearity and simplicity. We have been studying metric changes during growth and with lens-induced myopia in the chick eye. Thus, useful metrics must describe large changes in defocus as well as relatively smaller changes in image quality due to changes in higher-order aberrations. In order to test their behaviour, a large selection of image quality metrics were calculated for increasing amounts of defocus, ranging from 0D to 20D. They were also tested with increasing amounts of coma, an asymmetrical aberration, from 0.01µm RMS to 0.2µm RMS. With increasing amounts of defocus or coma and hence decreasing image quality, the majority of the image quality metrics examined varied almost monotonically. However, several metrics did not vary monotonically with defocus, some for good reason. These metrics were affected by the p phase shifts introduced by increasing amounts of defocus. In the case of the PSF, these oscillations corresponded to the absence of a bright central peak and a shift of intensity into an outer ring. Some metrics only varied monotonically in the presence of small amounts of defocus, but their behaviour was irregular with large amounts of defocus. Using these plots, the maximum equivalent defocus values over which each metric will be most sensitive to small changes in the aberrations was also determined. In summary, not all image quality metrics behave in a logical manner with the reduction in image quality associated with increasing amounts of defocus or coma. Some metrics will not perform well over a large range of image quality and their results should be considered with caution.C46