Animated Paintings

Abstract

The Game Technology Center (GTC) at ETH Zurich is currently researching augmented reality technologies to create novel and engaging interactive museum experiences within their ongoing project Behind The Art. The automated generation of animations for paintings of landscapes would be a very promising addition to the repertoire of augmented reality effects. This thesis aims to explore different new and existing animation techniques and their suitability to animate paintings.

We evaluated two different techniques of cloud animation. Simple horizontal motion as well as complex context sensitive motion prediction were combined to animate clouds within the back- ground. The motion prediction was also used to produce water animations and an appearance predictor was used to create a day-night cycle-like color transformation. These techniques are able to produce visually appealing animations as the results of our user study show. Four out of five participants generally prefer the animated versions over the static paintings. The results also indicate that smaller screens help to immerse the viewer by making small but disturbing imperfections less noticeable. Participants heavily invested in art and paintings tend to pre- fer the unedited paintings more frequently. Further work is needed to integrate such animated paintings into the Behind The Art platform and improve on certain aspects regarding the creation workflow and the water animations.