Traditional chroma keyers select one RGB (HLS, YUV, etc..) value and create the key signal using that selected color. This means that, when using a traditional chroma keyer, your incoming image should have exactly the same RGB values on every pixel. In real world, this is very unlikely to happen. Usually you will have some variance in your green screen image due to uneven lighting. In order to tolerate for the variances in green screen color, chroma keyers have to increase the tolerance (thus clipping fine details like hair) or increase the softness (introduces wrong transparent areas in the key).
Image based keying tries to solve this problem using a clean plate. This clean plate should be captured after you set your lighting and camera setups. An image based keyer would look for the changes between the clean plate and incoming video. This means that, we have individual key color setting for every pixel in our incoming image. So, we don’t need to add tolerance to clamp the uneven green values. As a result, we can extract fine details like hair, transparent object and contact shadows, even if the cyclorama is not perfectly uniform in terms of color.
This capability of image based keying allows you to lit your subjects more freely. You can create shadows on the ground floor using physical lighting and use that keyed shadow in your final composite.
Reality Keyer with moving camera
Reality Keyer uses clean plate as the source key color for every pixel. This is the way we suggest to key trackless camera subjects. When the camera moves, we need to create the clean plate every frame. This is achieved using the
node within Reality.
node is used to define your physical cyclorama geometry. After your cyclorama is defined and aligned with your camera tracking systems, this node can be used to generate the clean plates and soft garbage mask for your green screen production.
Basically, it is very similar to single frame clean plate capturing. After you set your lights and camera, capture a panorama of your cyclorama using the widest zoom level. The captured panorama is mapped on to the geometry and rendered every frame as a clean plate for the Reality Keyer’s clean plate input.