Interactive Insertion of Virtual Objects in Photos and Videos

Abstract

The introduction of virtual objects in photos or videos has been the focus of many Augmented Reality applications. This paper proposes a framework using image analysis methods to automatically detect scene features to introduce virtual 3D geometry objects. The current high-level features include surfaces, depth and scene orientation automatic detection. The main use of this technology is in AR tools, games or applications, which require the user to introduce an object that blends with a photographed or filmed scene. The main advantage of the proposed approach is that it can work only with one or two photos without prior knowledge, being ideal for mobile applications with camera or to be used with photo albums. Additional sensors can be added to increase reliability such as depth sensors or accelerometers but they are not essential. The main algorithms are intended to create a scene model in a few seconds and allow an interactive behaviour of the augmented objects.