Abstract
Natural images can have a very high luminance dynamic range, yet we know little about how the luminance range of a visual context influences perception. We used a high-dynamic range (HDR) display to measure how varying the luminance range of a grayscale Mondrian affects the perceived lightness of test patches. The Mondrians consisted of 5 × 5 arrays of square patches of fixed chromaticity. The central patch served as the test and the surrounding 24 patches provided the context. The Mondrians contained no segmentation cues indicating separate fields of illumination. The luminance ranges for the two contexts studied were 6900:1 and 31:1. The highest luminance patch was approximately the same for each context, and the surround patches were equally spaced in log-luminance. Observers matched the lightness of the central test patch to a palette of Munsell papers that was viewed in a separate illuminated chamber. Twenty-four test patch luminances were matched for each context. The range of test patch luminances spanned the range of the corresponding context. The palette matches allowed us to establish perceptual equivalence between test patch luminances across the two contexts, by equating the mean test luminance matched to each palette paper. Control experiments using uniform backgrounds indicate that our procedures reproduce salient features of classic measurements obtained with such backgrounds. The dynamic range of the context has a large effect. A 887:1 luminance test range presented in the 6900:1 context is matched to a 25:1 range presented in 31:1 context. This range compression, as well as the shape of the test-to-test mapping function, can be accounted for by models developed for lightness context effects obtained with uniform backgrounds: changing the stimulus range affects both a multiplicative gain and a subtractive offset. Much of the range compression may be attributed to a substantial change in the subtractive offset.
Supported by NIH RO1 EY10016 and NIH P30 EY001583.