Human Perception & Advancements in File-Based QC

Abstract

Archiving and retrieval of video picture assets has always presented an issue as to ensuring that the recovered asset contains the best quality picture possible. While current methods exist to detect (and correct) artifacts from the original camera negative, this is generally not the case with the medium available from the archive. The more common medium for retrieval is a digital video tape or film print. When retrieving video pictures from digital video tape or film prints, artifacts are generally introduced via methods that are difficult to detect without use of the human visual system, since many of these artifacts do not have a common, mathematically definable pattern to them. These artifacts can include film tearing, film dirt, analog noise, block-based digital drop outs and others. — This paper covers a newly designed metric and the implementation methods used to automatically find these types of artifacts without need of an external reference, substantially functioning and locating artifacts in the same method as the human visual system. The paper also shows the commercial viability of this metric in a system, and how this metric is useful and cost-saving for file-based content preparers compared to existing, manual processes for content review. The methods and system described herein are patent pending.