An empirical approach to visual display preference based upon modulation transfer function and luminance.

Abstract

Abstract : The purpose of the present study was to develop a three-dimensional preference space for displays as a function of the display modulation transfer function (MTF) and average display luminance. For any MTF-luminance combination, then, the goal was to generate a point in the third dimension denoting the preference for that pair. A paired comparison experiment was conducted where, on individual trials, observers viewed side-by-side images varying in MTF (5 levels) and average luminance (4 levels). The 5 X 4 = 20 combinations of MTF and luminance could be thought of as 20 filters. Preferences on individual trials were cumulated into empirical preference probability matrices which denoted the probability of preferring any one of the 20 filters over any of the 20 filters. A psychological model of preference, the Bradley-Terry-Luce (BTL) Model was then fit to the matrices in order to estimate a scale value or preference for each of the 20 filters or points in the three-dimensional space. Regression techniques (R(2)=.98) were used to generate a preference surface in the three-dimensional space, from which the preference for any display could be predicted. Additional analyses indicated that not only did ratings differ significantly based upon changes in MTF and luminance, but ratings change significantly (P<.001) as a function of using different scenery in the images. Finally, in predictive equations generated from the data, changes in MTF area (measured in percent contrast x cycles per degree of visual angle) tended to have about three times as much effect on preference as did changes in luminance (measured in footlamberts). Bradley-Terry-Luce model, Image quality, Multidimensional scaling, Contrast, Modulation transfer function, Paired comparisons, Gaussian MTF, MTF, Resolution.