Book Description

This dissertation, "Unified Percepts in Three-dimensional Space Derived From Motion in Depth or Rotation in Depth" by Chak-pui, Terence, Lee, 李澤沛, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: ABSTRACT A long-standing challenge to understanding the human visual system is to reconcile our unified visual experience of a three-dimensional (3D) world with findings suggesting that the visual system operates, at least partially, as a modular system. This dissertation is composed of studies that investigate how systematic changes in retinal image structure, and changes in binocular disparity arising from either motion in depth or rotation in depth, are used to derive a number of perceptual judgments in 3D space. The assumption that looming and stereomotion mechanisms operate in tandem, and feed into an early joint motion-in-depth mechanism, is challenged by the findings of Chapter 3. Using the flash-lag paradigm, it was found that stereomotion, but not looming, induces an apparent displacement in the position in depth of a flashed object, and the effect is speed dependent. This finding suggests that, in terms of the encoding of metric depth, the two mechanisms produce qualitatively different results. Moreover, the flash-lag-in-depth phenomenon is accompanied by a novel size effect, which appears to contradict the geometrical predictions typically used to account for size illusions associated with depth perception, and serves as an evidence of top-down processing in deriving congruent estimates of image size and position in depth. Assuming that different mechanisms underlie the computations of 2D motion and motion in the depth, their outputs must be combined at some stage, given that objects in the 3D world often traverse frontoparallel and median planes simultaneously. Findings of how 2D motion affects the apparent speed of motion in depth are reported in Chapter 4. Dots defining a frontoparallel square conducted random walks on its surface, while the square itself moved towards the observer. It was found that apparent speed in depth increases as a function of the speed of random 2D motion. Additionally, the visual system appears to combine motion signals in the twoplanes by vector summation when motion in depth is defined by stereomotion, whereas averaging is implemented when motion in depth is defined by looming. The detection of 3D structure often requires the integration of local motion signals traversing 3D space. An investigation of this process, using cylinders with an apparently identical 3D shape, is reported in Chapter 5. A cooperative integration of disparity and structure from motion was found: compared to bi-rotation in depth, a unified direction of rotation in depth improves detection performance, which can be explained by the functional properties of MT neurons. While the apparent 3D shape of cylinders remains constant for a range of dot densities, sensitivity to the existence of cylinders embedded in random noise increases as a function of dot density, which implies a gradation in 3D-structure interpolation. To examine the effect of apparent orientation in 3D space imposed on the bi-stable appearance of rotation in depth, in Chapter 6 the detection of spontaneous and physical reversals in the direction of motion were compared for cylinders and Necker cubes defined by structure from motion. The frequency of spontaneous reversals occurring with the Necker cube is far lower than with the cylinder and with Necker cubes defined by outlines. Furthermore, physical reversals that occurred with the Necker cube are much more detectable than with the cyli