While Augmented Reality extends the user’s environment by virtual objects, Diminished Reality removes real elements out of the user’s view. Vision-based techniques detect undesired image elements and replace them by other image content providing an overall consistent impression for the observer. Diminished Reality is realized in real-time for video streams allowing for arbitrary object removal in live video broadcasts. The image manipulation typically cannot be recognized by the observers.

We investigate in different computer vision based approaches for realizing Diminished Reality systems. During our research we already developed the world-wide first real-time capable Diminished Reality system. The technique can directly be applied in almost every environment without any previous knowledge. Further, our system runs on mobile computers such as laptops or tablets. The user may interactively select undesired objects on a live camera image. These objects will then be removed by the Diminished Reality technology in the live video stream. The following video demonstrates some application areas of our technology.