Automated Edge De-Spill For Faster and More Efficient Chroma Key Compositing in Visual Effects

In the visual effects compositing industry, the laborious manual process of convincingly merging matted elements via chroma-key (blue or green screen) to background plates has remained mostly un-automated. The bulk of this process is manually treating the edge around the foreground keyed object or character to match the perimeter color of the subject, thus removing any halo of conflicting color in the motion blurred or out of focus regions. This manual rotoscoping and edge treatment process takes up the artists’ precious time and, due to this, eats into the budgets of most struggling VFX houses today. The De-spill Edge Hack toolset is an automated toolset that uses stepped un-pre-multiplication and blur procedures to extend the edge color of the foreground element to fix edge color continuity and remove edge pollution caused by the chroma key backdrop (blue or green usually). The toolset updates per frame leaving little or no need for the old school manual approach, not only saving time was ensuring accuracy of edge match to localized edge interior perimeter colors. The following study looked to analyze and deduce whether the toolset is good enough for use in production via a set of demonstrations and surveys with feedback from anonymous participants. A video demonstration of the toolset was given as well as side-by-side “toolset vs manual approach” multiple choice to deduce which technique would yield a better de-spilled edge. The sample size was 57 participants with a knowledge range of little to 6+ years’ experience in the field of visual effects. The overall findings showed that participants preferred the tool 22% percent of the time while 48% preferred the manual approach. The remaining 30% could not tell the difference. Further comparisons of data between 6 categories of changed variables for the 12 side-by-side also showed similar ratios of choice. Only small hints of possibly one of the six variables showed an impactful change in the ratios. Participants were given 5-point Likert scores (1-Strongly Agree, 5 Strongly Disagree) on whether they thought the tool was production ready and if the toolset needed additional control features. There was a mixed but slightly overall agreement (46.6% agreed) that the tool was production-ready with most participants desiring additional control features (88.2% strongly agreed/agreed). The written suggestions from participants were positive overall with a few folks asking for light-wrap additions to the toolset. Creating an automated de-spill workflow that can handle common difficult composite edges would be crucial in speeding up production workflow and edge accuracy.