PREDICTING READABILITY OF TRANSPARENT TEXT ON TEXTURED BACKGROUNDS

L. F. V. Scharff

Stephen F. Austin State University, Nacogdoches, TX

and

A. J. Ahumada

NASA Ames Research Center, Moffett Field, CA

Purpose: The aim here is to find measures that predict the effects of background contrast variations on text readability. Previous work showed that a combination of the text contrast and the background contrast energy provided a useful measure (Scharff & Ahumada, Optics Express, 2000). We wanted to know whether this measure was equally useful when the text was combined transparently (additively or multiplicatively) with the background.

Methods: Readability was measured by accuracy and search latency as in the above reference. Observers were 54 undergraduates. There were three experimental factors: text contrast (0.30, 0.45), transparency combination rule (additive, multiplicative) , and masking pattern (uniform, “wave”, “culture”). The two patterns were used in prior experiments and were originally obtained from a website providing backgrounds for web designers. They were adjusted to have the same mean luminance, but they had different RMS contrasts (0.27 and 0.15, respectively).

Results: Performance was significantly worse for the low text contrast, the additive combination rule, and the “culture” pattern. (Both patterns were worse than the uniform background.)

Conclusions: Our original measure predicts no effect of combination rule and more masking by the “wave” pattern. It was borrowed from models for signal detection, where the effect of the signal on masking and adaptation can be ignored. Even though the text affects fewer than 20% of the pixels, this data is fit much better by using both the text and the background to compute the text contrast and the masking RMS contrast. The adjusted measure correctly predicts the direction of the effect of the combination rule and predicts a much smaller effect of the background pattern RMS contrast.