Region-of-Interest-Based Subtitle Placement Using Eye-Tracking Data of Multiple Viewers

We present a subtitle-placement method that reduces viewer's eye movement without interfering with the target region of interest (ROI) in a video scene. Subtitles help viewers understand foreign-language videos. However, subtitles tend to attract viewers' line of sight, which cause viewers to lose focus on the video content. To address this problem, previous studies have attempted to improve viewer experience by dynamically shifting subtitle positions. Nevertheless, in their user studies, some participants felt that the visual appearance of such subtitles was unnatural and caused them fatigue. We propose a method that places subtitles below the ROI, which is calculated by eye-tracking data from multiple viewers. Two experiments were conducted to evaluate viewer impression and compare line of sight for videos with subtitles placed by the proposed and previous methods.