Visual Search Work (3-D Visual Search)

My work in this area has incorporated depth processing in two ways: through the use of display which possess stereoscopic depth cues and and those which possess monocular depth cues.

Return to Scharff research summary page.


How are visual searches processed across three dimensions (stereopscopic depth)?

Most previous research investigating visual search has placed targets and distractors in a single depth plane. However, in our everyday lives, we make searches across three dimensions.

As presented at ARVO 2000, my most recent research further indicates that spreading the distractors in a visual search across multiple depths planes speeds the search times. More specifically, it seems that the more depth planes used the faster the search (5 subjects performed searches with a single depth plane, 3 depth planes, and "cloud" conditions where the distractors were spread among at least 9 depth planes). This benefit occurs even when the participants were not primed to the depth location of the target.

Scharff, L.V. (2000). Serial Visual Searches Processed More Efficiently Across Multiple Depth Planes. Investigative Ophthalmology and Visual Science, 41, S424.  [Abstract]

See example stimuli.

Earlier Work:

A preliminary study (with Heather Melton) using two depth planes was presented at the Southwestern Psychological Association meeting in April 1998. A second study, using three depth planes and a modified stimulus was presented at ARVO in May 1998. Both studies indicate that conjunctive searches are processed more efficiently across multiple depth planes. Prior knowledge of target depth plane does not seem to systematically speed the search.

With then undergraduate student Aaron Aiza, we compared 2-D and 3-D searches using one and two targets defined by luminance. Findings showed that two targets defined by luminance were detected more slowly than one target defined by luminance. Although two targets were detected more slowly, they were still processed preattentively. Two other experiments were created in which targets and distractors were placed across three depth planes (target location was either blocked or random). The analysis showed that the 3-D visual searches were preattentive and faster than the 2-Dsearches for two targets. This suggests that the human visual system can simultaneously process more than one disparity channel. In a comparison of the two conditions (random v. blocked), the results showed that prior knowledge of target depth (blocked condition) leads to faster searches than if the target's depth location is not known (random condition). This suggests that the human visual system can selectively attend to different disparity channels. This work has been published in the Proceedings for the National Conference of Undergraduate Research, 1999.

 

Visual Search Using More Naturalistic Backgrounds and Dimensionality Cues

The earlier work with Tisha indicated that the implied dimensionality may not be what allowed the cubes in Rens and Ensink's (1990) study to be preattentively processed. We found that implied lighting was from the top or bottom, then the searches are fast and parallel (as in Enns and Rensink's study). In contrast, for left/right lighting, searches tended to be serial in style. However, the cube shape that was used by myself and Tisha was different than that used by Enns and Rensink -- it is possible that the unequal surface areas led to the difference in findings. Therefore, Amanda Collier and I further investigated the issue using both cube types and implied lighting from the top, bottom, left and right. A summary of this work was presented at the South Western Psychological Association meeting, April 2000.

With graduate student Tisha Reid: we investigated visual search using targets / distracters which incorporates monocular depth cues and which were placed on complex backgrounds. Results from this work should be more generalizable to naturalistic visual search tasks than the traditional visual search experiments using blank backgrounds have been. A summary (poster session handout) of the main experiment was presented at ARVO 1997.