This research was conducted to achieve four objectives. The first objective was to develop an analytic model to predict the expected luminance distribution through the shadow mask structure on a color CRT display system. The model incorporates functions to describe the unique features of a color CRT, that is, the discrete sampling imposed by the shadow mask/ phosphor-dot arrangement as well as the electron beam phase relationships. The model also includes a flexible beam profile which allows the user to specify the desired shape of the beam profile, that is, whether the profile is described with a Gaussian, leptokurtic, or platykurtic distribution. This objective was fully satisfied with a computer program written in Lightspeed C which runs efficiently on Macintosh computers.

The second objective was to determine detection thresholds for various levels of misconvergence of the three electron guns. When the three guns are properly registered, the luminance profiles converge and one perceives a color combination rather than the separate red, green, and blue luminances. Misconvergence is perceived by a change in the overall color or by color fringes, for example, a red edge to a yellow line. Past research has shown that threshold detection of misconvergence occurs when the primary beams are misconverged by 1 to 2 visual arcminutes of separation. This finding was replicated in this research for the two-color beam combinations which have previously been investigated, as well as for a white pixel, which involves all three guns.

The third objective was to demonstrate the effect of misconvergence on the performance of a visual task and on subjective estimates of image quality. While subjective quality and threshold detection have previously been investigated for some color combinations, the three tasks (i.e., threshold detection, visual task performance, and subjective estimates) have not been systematically combined within the same data set for a variety of misconvergence conditions. This research provides such a composite data set. The subjective quality estimates were significantly correlated with the threshold detection data. In other words, as misconvergence of the display image increased, the probability of detection of misconvergence increased and the subjective quality rating decreased. However, the selected visual task (a short reading task with average reading time of 6.5 s) was not significantly affected by very large levels of misconvergence. Rather than conclude that the levels of misconvergence used in this research do not affect reading task performance, a more comprehensive visual task (e.g., a longer editing task, a random search task, or a map reading task) should be evaluated.

The final objective was to evaluate the ability of selected image quality metrics which are computed from the model to predict threshold detection, subjective quality ratings, or visual task performance. The three metrics computed in this model (MTF Area, MTFA, and SQRI) are all based upon the modulation transfer function (MTF) of the display. These three computed metrics were for all practical purposes constant across the range of misconvergence. While this result was unexpected, it does suggest (1) that a model based only on luminance may be deficient because of the omission of chromaticity, and (2) that MTF-based metrics may not be an appropriate representation because misconvergence does not change the display’s ability to transmit information, but is a phase shift along the shadow mask.

As summarized, this research successfully met three of the stated objectives. Further, it points toward future research opportunities to further this type of modelling effort and to successfully develop image quality metrics for color displays.