Preferred levels of image sharpness and their relation to image structure in thermal dye transfer prints

Abstract

The objectives of this research were (1) to develop scales of preferred image sharpness for thermal dye transfer prints and (2) to relate these scales to an objective measurement of image quality.

Interval scales were developed using two different indirect scaling techniques, paired comparison and rank ordering, for three groups of participants (novice photographers, advanced amateur photographers, and Kodak image quality experts). Differences between scales developed through the separate scaling procedures were minor, suggesting that either procedure would produce a reliable and valid interval scale. Given the time required to perform the method of paired comparisons, the method of rank ordering would likely be the preferred procedure.

The results from multiple range tests of the interval scales found that higher levels of image sharpening (filter levels 1.5x to 2.5x) were generally preferred. However, for some conditions, particularly the portrait, lower levels of sharpening were generally preferred. While it appears that the preferred level of image sharpness may be somewhat dependent upon scene content, participants always preferred a small amount of sharpening to none at all. For all conditions examined, stimuli that received no sharpening were the least preferred.

In relating interval scales to an objective measure of image quality, scale values developed from within a level of addressability were highly correlated with values of modified MTFA for all participant groups. These correlations show that a strong relationship exists between preferred levels of image sharpness and increasing levels of luminance modulation for edges in thermal prints. However, values of modified MTFA were not well correlated with interval scale values developed from experts participants for the between levels of addressability portion of the experiment. Values of MTFA were considered modified due to limits that were imposed on the bounds of integration. These results suggest that the preferences of experts may not represent those of the average consumer for desired levels of thermal print sharpness.