Abstract
Focus cues—blur and accommodation—are regarded as very weak depth cues. The research that led to that assessment has, however, been greatly limited by an inability to present focus cues in a fashion consistent with natural viewing. We recently developed a novel volumetric display that allows the presentation of near-correct focus cues along with standard depth cues like binocular disparity. We used the display to re-examine the usefulness of blur and accommodation as cues to depth. We created stereograms with anisotropic textures that created disparities specifying a disk in front of a background. Probability-based theories of cue combination predict that as the disparity signal becomes less reliable, the percept should be more heavily determined by the focus cues. We varied the reliability of the disparity signal by changing the dominant orientation of the texture. Reliable (texture vertically oriented) and unreliable (texture horizontally oriented) disparity signals were presented in three conditions: i) focus cues specified zero depth, as they do in conventional 3d displays; ii) focus cues and disparity specified the same depth, as they do in natural viewing; iii) focus cues specified more depth than disparity. We presented stimuli in two intervals and observers reported the interval with more apparent depth. As expected, when focus cues specified zero depth, subjects saw less depth when the disparity signal was unreliable than when it was reliable; when focus cues specified more depth than disparity, the effect was reversed: subjects saw more depth with unreliable disparity. Based on pupil-diameter fluctuation, accommodative fluctuation, and optical aberrations, we computed the theoretically expected depth-from-focus likelihood functions. They are reasonably similar to the likelihood functions estimated from the data. These results show that focus cues provide a metric depth signal that is combined in a statistically reasonable fashion with disparity.
William C. Ezell Fellowship, NIH and NSF.