The initial purpose of video captions were to convey audio content to people with hearing loss. However, captions are beneficial to many other user groups (e.g., second-language learners, viewers in public spaces). The two most common methods for ensuring captions visibility are: to use plain white text (sometimes with an outline) or to place the caption within a darker (sometimes black and opaque) box. Background videos with overly bright spots can reduce legibility of white text. Though a dark box around the text guarantees a high degree of visibility, it obscures a larger section of the video. To improve visibility, one would have to ensure high contrast without additional obstruction of the visual content.

I set out to evaluate the efficacy of dynamically determining the colour of captions based on the surrounding area to ensure high contrast using a fuzzy expert system. Previous work leveraged a fuzzy expert system to determine a single contrasting colour selection for captions; I implemented a fuzzy expert system that determines different colours along a black-white gradient to select the colour for parts of the caption depending on the background. Along with subjective analysis of the resulting colourations, I used peak signal to noise ratio to compare the traditional solid colouring and the proposed fuzzy colouring. I was able to develop a promising proof of concept for dynamically adaptive captions. Next steps include adjusting the ruleset for the expert system to optimize the colour selection. Furthermore, future work could involve conducting usability studies with human participants to test readability and preference. Overall, my proof of concept appears to be a positive step towards improved accessibility and usability of video captioning.