Camera Tracked 3D Animations

Tracking the video

Firstly we stabilize the video. This process, while taxing on a workstation, doesn't actually require much human time - just hours of CPU time. All the the bumps and jumps from the footage are removed and the result is silky smooth video. If the shot is god then less stabilization is needed, but if it was taken from a small helicopter or in turbulent weather then more is needed and there is a slight reduction in finished quality.

The actual tracking of the video is partly automated task although still requires a great deal of human time. The simple fact is that it is a very complex process that cannot be done by man or computer alone†. Points in the video are marked and through a process of manual input and automated tracking the computer determines a set of moving 3D coordinates that depict where the camera was at each point in time. As you may imagine, especially if you have a mathematical background, the computational power to determine this and track the points is somewhat huge and would have been next to impossible in the recent past.

The process of tracking a few minutes of video footage can take days, possible weeks, depending on the quality of the shoot and the subject.

Adding the 3D animation

Building the 3D model itself is a different process again, so from this point on we will assume it is done. The camera coordinates are brought into the 3D scene as is the stabilized footage. Then there is another painstaking process, which is aligning the camera and the 3D model with the video background. Even though the footage is tracked the coordinates will not actually relate to the 3D scene until adjustments are made. This may take hours or days to get right.

Then the lighting in the 3D scene needs to be adjusted to match the lighting on the actual day the video was shot. The idea is to get a seamless combination of the model and reality.

Other factors come in to play as well, such as foreground objects in the video. These require a technique called rotoscoping, which is a moving layer of transparency that is added to the final 3D rendered footage to ensure that anything that is supposed to be in front of the building is still there after the 3D footage has been added. Once again this is a computer/human combination that can take some time. Normal rates for Archiform 3D to create camera tracked 3D animations do NOT include rotoscoping and we will shoot to avoid it's need, but if you require it we can do it for an additional fee.

Part 2 of 3

Next Uses for tracked 3D animations, Sample movies >

< Back Introduction, Shooting the video footage

†Humans could do it but the process would take way too long