Abstract

We measured resolution of edge (the line) blur as a function of edge (and line) contrast. Previous attempts to predict resolution thresholds involved many modeling assumptions concerning the visual system’s sensitivity and/or the retinal sampling density. Our goal is to avoid assumptions by measuring the visibility of those multipoles that are used in the resolution task. The task of discriminating edge (or line) blur is equivalent to the task of detecting a dipole (or quadrupole) test pattern in the presence of the edge (or line) pedestal. The multipole test threshold can be related to the variance of the blur by using the relationship $P_{v+\epsilon }(x)=P_v(x)+0.5\epsilon P_v^{\prime \prime }(x)+O\left({\epsilon }^2\right)$, where P_v(x) is an edge or line with a blur of variance v and where ϵ (with units of min²) is the ratio of test threshold to pedestal strength. A sample result is that a dipole of 4 % min² strength was visible on a 100 % contrast edge pedestal, corresponding to a blur variance of 0.04 min² (equal to a standard deviation of 0.2 min or a 0.28-min separation between a pair of lines). This small resolution threshold would be surprising in terms of the retinal sampling density but is not surprising in terms of our test-pedestal approach.

© 1988 Optical Society of America