Keesee, Robin Lee2023-04-072023-04-071975http://hdl.handle.net/10919/114429Determination of some form of modulation detectability threshold for sinusoidal gratings is a prerequisite to use of most measures of image quality for line-scan imaging systems. To develop predictive models for these thresholds, an experiment was conducted to determine the modulation detectability threshold functions for a range of system parameters typical of medium- to high-resolution low-light-level television systems. Specifically, the psychophysical method of adjustments with target grating modulation as the dependent variable was employed in an experiment where, at each of 10 combinations of video line rate and video noise passband, each of 7 subjects received 2 replications at every combination of 10 spatial frequencies, 2 target orientations, and 5 noise amplitudes. Line rates were 525, 945, and 1,225 lines per frame; the spatial frequencies were between 1 and 20 c/deg; and the gratings were oriented vertically (perpendicular to the video raster) and horizontally (parallel to the raster). As the three highest spatial frequencies were not used at the 525 line rate, horizontal orientation, there were 13,160 trials. The target gratings were displayed in a 6.375 in. (16.2 cm) square center portion of a 10 in. by 14 in. (25.4 cm by 35.6 cm) video monitor with both the grating and surround having a mean luminance of 15 fL (14 mL). Viewing distance was 40 in. (101.6 cm). Using a step-wise regression technique to evaluate candidate variables, a model predicting the modulation threshold as a function of the other variables was developed for each grating orientation. In the vertical orientation, a correlation coefficient of .77 was achieved between the experimental data and a three-variable equation consisting of a variable representing the integration of noise power from zero spatial frequency to the grating spatial frequency, the value of the modulation threshold determined by DePalma and Lowry (1962) at the grating spatial frequency, and a variable representing the noise power in the spatial frequency region of the grating. The addition of twenty-two other variables increased the correlation to .82. For the horizontal orientation, a similar correlation of .74 was obtained with a three-variable equation including the variable representing an integration of noise power from zero to the grating spatial frequency, the inverse of the raster spatial frequency, and a shaping variable equal to the square of the grating spatial frequency minus 9 c/deg . The inclusion of twenty-three additional variables yielded a .81 correlation coefficient. The correlation coefficients are between the models and the complete data sets, not means across subjects and replications. These models are useful in the application of the modulation transfer function area and other spatial frequency-based line-scan image quality metrics.iv, 196 leavesapplication/pdfenIn CopyrightLD5655.V856 1975.K433Dissertation