Search

This dissertation research had three primary objectives. The first was to develop and evaluate a metric of image quality that incorporates a model of suprathreshold contrast and is based upon the perceived magnitude of suprathreshold contrast. The second objective was to determine the effects that common display characteristics, such as MTF shape, noise, and field of view, have on a display observer's ability to perceive egocentric motion from a display. The third and final objective was to provide a discussion of the use of image quality metrics for the evaluation of displays that are designed to facilitate motion perception. To meet these objectives, two experiments were conducted. The first experiment investigated the effects of the display Field of View, Dynamic Contrast Range, Noise Level, and Bandwidth on perceived image quality. The second study investigated the effects of these same display parameters as well as the effect of terrain type on observers’ sensitivity to changes in the speed of image motion. Existing visual psychophysical data were used to propose the two additional image quality metrics, the Perceived Contrast Magnitude (PCM) and the Weighted Perceived Contrast Magnitude (WPCM). Each of these metrics provides some additional model parameters to the Modulation Transfer Function Area (MTFA). However, based upon the studies that were conducted in this dissertation, it appears that the Integrated Contrast Sensitivity (ICS) and Square Root Integral (SQRI) provide significantly better estimates of perceived image quality than do MTFA, PCM, or WPCM. Results indicate that the display parameters of Bandwidth, Field of View, and Noise Level, plus several interactions significantly influenced the observers’ sensitivity to changes in the speed of images. One interaction indicated that observers were more sensitive to changes in the speed of images when the scene contained a lot of contextual information or when the scene was viewed through a display with reasonably high bandwidth and low noise level. However, when the bandwidth of the system was decreased, the noise level was increased, and the contextual information in the image was reduced, the observer's sensitivity to changes in motion was degraded when viewing the image through a small field of view display (20 deg) but it was not degraded when viewing the image through a larger field of view display (47 deg). Based upon the evidence collected in this dissertation, it appears that displays designed to support motion perception should have a relatively large field of view with partitions to facilitate relative motion perception. The factors of Dynamic Contrast Range and Field of View were shown to have different effects on the perception of image quality than they have on the perception of changes in the speed of an image. Therefore, the use of image quality metrics for the evaluation and design of displays that must support the veridical perception of motion should be questioned. Instead, it appears that a task specific evaluation technique should be developed for the evaluation of these displays.

An improved feature geometry based, low-cost, machine vision software system was developed by conducting a statistical reliability study of a previously developed system, evaluating alternative methods of completing the machine vision process, developing more effective edge linking and corner detection algorithms and implementing these new algorithms into a more efficient and reliable version of the existing machine vision software. The resulting system utilizes the same machine vision hardware that was utilized by the previous machine vision software system. This hardware includes an Apple IIe microcomputer, that serves as the host computer for the machine vision system; a 4K frame buffer, that is used to capture designated areas of the image obtained by the image sensor; a General Electric TN2500 Digital Camera, that serves as the sensor; a monitor and a Microbot robotic arm that is used to pick up and move an object for demonstration purposes. The resulting system is capable of recognizing a flat surfaced, randomly positioned rectangular object that contrasts well with its background within about 11 seconds with an overall reliability of about 80%. This system is equipped with an interactive user interface that allows the system to be operated and maintained by a person with very little knowledge of computers and machine vision. The interface gives the user the capability of changing parameters that affect how the machine vision system operates. These features combined with the system’s ability to interactively output the data that is generated during the system’s operation makes this system a valuable teaching tool to demonstrate the application of machine vision to students.