Abstract
Our previous work showed that a combination of text contrast and background contrast energy provides a useful metric for predicting text readability on a textured background (Scharff, Hill, & Ahumada, Optics Express, 2000). More recently (Scharff & Ahumada, Journal of Vision, 2002), we showed that a version of this metric predicted readability using different types of text transparency (additive vs. multiplicative text combinations with the background). The measure used both the text and the background to compute the text contrast and the masking RMS contrast. In that study, there were three experimental factors: text contrast, transparency combination rule (additive, multiplicative), and masking pattern (uniform, “wave”, “culture”). The metric had two problems, it only partially predicted the effects of transparency and it incorrectly predicted more masking by the masking pattern with the higher root-mean-square contrast. Here we report measurements of participants' ability to identify the individual letters cut out with their backgrounds that were presented in the earlier (2002) study. Because of the large number of letters, half of the 30 participants were assigned the additive conditions and half the multiplicative conditions. An analysis of variance of the latency responses shows all three factors to be significant with no significant interactions. The letter identification latencies were longest for the masking pattern with the higher root-mean-square contrast as predicted by the metric and were longer for the positive contrast additive letters than for the negative contrast multiplicative letters. When the contrast gain of the metric for negative contrast is set to twice that for positive contrast, the rank correlation of this asymmetric contrast metric with the average latencies was 0.986.