Text legibility in augmented reality with optical see-through displays can be challenging due to the interaction with the texture on the background. Literature presents several approaches to predict legibility of text superimposed over a specific image, but their validation with an AR display and with images taken from the industrial domain is scarce. In this work, we propose novel indices extracted from the background images, displayed on an LCD screen, and we compare them with those proposed in literature designing a specific user test. We collected the legibility user ratings by displaying white text over 13 industrial background images to 19 subjects using an optical see-through head-worn display. We found that most of the proposed indices have a significant correlation with user ratings. The main result of this work is that some of the novel indices proposed had a better correlation than those used before in the literature to predict legibility. Our results prove that industrial AR developers can effectively predict text legibility by simply running image analysis on the background image.

The application of augmented reality in industrial environments requires an effective visualization of text on a see-through head-mounted display (HMD). The main contribution of this work is an empirical study of text styles as viewed through a monocular optical see-through display on three real workshop backgrounds, examining four colors and four different text styles. We ran 2,520 test trials with 14 participants using a mixed design and evaluated completion time and error rates. We found that both presentation mode and background influence the readability of text, but there is no interaction effect between these two variables. Another interesting aspect is that the presentation mode differentially influences completion time and error rate. The present study allows us to draw some guidelines for an effective use of AR text visualization in industrial environments. We suggest maximum contrast when reading time is important, and the use of colors to reduce errors. We also recommend a colored billboard with transparent text where colors have a specific meaning.