The goal of this project is to build a system, which is able to create a continuous tracking shot from separate already filmed fragments. These fragments do not cover the whole range of the camera movement in the final video, so synthetic sequences have to be generated, which seamlessly fit between the end of one and the start of the next fragment.
The following specific example illustrates the intended application of the system: one wants to take a shot, in which the camera glides from above the roofs of a city, down towards a building, through a window, and continues inside the apartment into a room, finishing with a close-up of the main character. A shot with a crane, starting over the city and moving close to the window can be realized, as can a shot moving through the apartment using a steadycam or dolly. But the sequence through the window has to be computer generated to connect the two different sequences taken by different cameras. Our solution uses additional still images, taken in between the fragments. These images do not have to lie on the final camera path. By using the video fragments and still images, a partial reconstruction of the scene is generated, which allows free virtual camera movements in the area covered by the images.
A coarse real-time preview will be used on the set, to determine if the video fragments and still images contain enough data to synthesize the missing video sequences. Off-line processing in post production will generate the final shot, with a maximum of leeway for corrections of the camera movements.The audience should not realize that it sees a combination of three different shots, one of them even taken by a virtual camera. Therefore, all the visual quality and vividness has to be preserved in the digital images. The virtual and the real shots have to match exactly in motion and in color, and have to be blended carefully. Since most of the productions are documentaries, the images have to stay authentic. One cannot – as it is usually done for feature films – just invent the scenery in order to adapt it to the technical needs.The format of the resulting, merged video should be broadcast-quality high-definition, which requires much higher resolution and image-quality than broadcast PAL or DVD targeted applications.
DragonCAM will be a "virtual film-shot generator" achieving the high quality standards in today’s film making. In combination with the existing technologies of FWG – Brains & Pictures, dragonCAM will combine the advantages of real-life film-shots and the possibilities of computer generated film shots.