Abstract
Background: Several visual “cues” can be used to discriminate surface materials: color, lightness, roughness, gloss, texture, and so forth. The visual system can potentially perform better in discriminating materials by integrating information from several such cues. In real-world conditions, however, there might be considerable uncertainty about the most informative cues. We tested the integration of color and gloss while manipulating extrinsic uncertainty in a discrimination task.
Methods: We used spectrally rendered images of 3D shapes as stimuli. Diffuse and specular reflectance were independently varied to produce a two-dimensional stimulus space, with stimulus color varying from blue to green and specularity from matte to glossy. On each trial, the observer compared a reference and a test stimulus, which in different conditions differed from each other in color, gloss, or both. The observer identified the bluer/glossier stimulus. Different conditions were either blocked (no uncertainty) or mixed within a block of trials (uncertainty). We measured a discrimination threshold in each condition.
Results: Performance was better with two cues compared to single cues. We fit models based on signal-detection theory to the data from each condition, assuming independent processing and optimal integration of the two cues. The models gave good fits to both no-uncertainty and uncertainty conditions.
Conclusion: The human visual system can integrate information from color and gloss to improve material discrimination. The model fits to both no-uncertainty and uncertainty data indicate that observers are able to flexibly change their strategy according to the task structure.