Deep Compositing

Abstract

At Animal Logic, we exploit the potential of images produced by 3d rendering rather than physical cameras to improve pipeline and workflow processes in compositing, in particular facilitating the in- tegration of stereoscopic and volumetric elements. Whilst tradi- tional compositing happens between flat images, many operations depend on the depth relationship of image elements. We render and composite with “deep images”, which contain information about coverage and depth along each pixel. This enables artists to achieve standard compositing operations without manual, explicit image segmentation or layering as well as decoupling rendering and com- positing by obviating holdout mattes. Post-render application of depth-of-field may also be approximated with fewer artefacts.