Abstract
Luminance can vary widely across a visual scene, by up to 106:1, requiring multiple luminance normalization mechanisms to support visual acuity. Cortical mechanisms underlying brightness perception and acuity are typically studied via laboratory stimuli at ~100:1 luminance contrast ratio (‘standard dynamic range’, SDR). How the cortex processes scenes with contrast ratios over 1000:1 (‘high dynamic range’, HDR) is poorly understood. We hypothesized that, because visual neurons encode both luminance and orientation, luminance and orientation processing may interact non-linearly during visual recognition depending on luminance dynamics. We measured EEG, eye tracking, and visual discrimination behavior under a two-alternative forced choice (2AFC) task. Stimuli consisted of 45- and 135-deg Gabors presented on a 5 × 5 grid of luminance patches (100:1 patch contrast, 10000:1 peak contrast including Gabors). The target was a contrast blend of Gabors at the two orientations, and subjects indicated via keypress the orientation with the higher contrast. In one condition (‘similar’), the co-oriented flanker patches were similar in luminance to the target patch, and in the other condition (‘brightest’), the co-oriented flanker patches were the brightest patches. Dependent variables included behavioral response time and accuracy, stimulus and ocular-locked EEG amplitude, latency and frequency, and pupil size. Our results showed that under decreasing luminance, responses are biased towards the orientation of flankers that are similar to the target luminance (‘similar’ condition), suggesting a facilitation effect that is luminance-context-dependent. The facilitation effect was abolished under the ‘brightest’, increasing or stable luminance, and SDR luminance conditions. Under stable luminance and uniform flankers (same luminance and orientation), responses were biased away from the flankers, consistent with assimilation of co-oriented target and flanker Gabors and ‘pop-out’ of the orthogonal target Gabor. These results suggest that opposing luminance-orientation contextual mechanisms significantly bias orientation discrimination behavior, depending on dynamic luminance normalization and on surround uniformity.
Acknowledgement: Research was sponsored by the Army Research Laboratory and was accomplished under CAST 076910227001 and ARL-74A-HR53