Computational Aesthetics: Improving Photograph Composition by Detecting Distractions

摘要

Background distraction is one of the common photographic composition problems that affect the photograph visual appeal according to our survey. In this paper, we proposed a method to detect distractions by combining subject detection and human visual attention detection based on a bottom-up visual attention model. For photographing a single person as subject, our method can detect the subject based on a single image. After that, our method detects background distraction, and provides the user a suggestion for a better composition. The whole process is fully automatic. Results show that the better-composed photographs generated by our method are close to professional photographers' suggestions. Our method is fast and has the potential to be incorporated into a digital camera firmware, so that the camera has the function of directing the user to better compose their photographs on-site.