Abstract
Most surfaces reflect light from external sources, but others emit their own light, or glow. Glowing surfaces often signify an important feature in the environment (e.g., heat source or bioluminescent life form), but we know little about how the visual system identifies them. Here, we show that perceived 3D shape is critical for perceived glow. In Experiment 1, we created "dark-means-deep" stimuli by rendering stereoscopic pairs of wavy 3D surfaces under diffuse light (non-directional lighting on a cloudy day). This generated stimuli with bright peaks and dark valleys. We created "bright-means-deep" stimuli by using the same luminance images, but reversing the disparity to get dark peaks and bright valleys. Subjectively, dark-means-deep stimuli appeared evenly lit from the front, whereas bright-means-deep stimuli produced a vivid impression of glow. On a mirror stereoscope, we displayed dark-means-deep and bright-means-deep stimuli side-by-side, and asked observers to choose the one that appeared to glow. Five of six observers consistently identified the bright-means-deep stimuli as glowing. In a follow-up study, we assessed depth percepts for the same observers and stimuli using a depth-probe task ("is the probe on a peak or in a valley?"). Five of six observers performed almost perfectly. The sixth observer (the anomalous observer from above) ignored disparity cues, always judging bright regions as peaks and dark regions as valleys. Thus, this observer was anomalous because disparity cues did not affect their shape percepts. In Experiment 2, we used the same observers, stimuli, and methods, except that we used motion parallax instead of disparity to reveal surface relief. All observers, including the anomalous one, identified bright-means-deep stimuli as glowing, demonstrating that the glow effect was not tied to a particular depth cue. Our results demonstrate that human vision has a sophisticated understanding of lighting geometry, interpreting complex shape-luminance relationships to identify glowing surfaces.
Meeting abstract presented at VSS 2016