Researchers from the University of Washington’s Graphics and Imaging Laboratory are developing video enhancement software which utilizes data from related still images. Currently only for use with ‘still-life’ scenes, the process incorporates information from higher resolution photos and automatically maps it to the corresponding areas of the video in need of improvement – increasing resolution, dynamic range, ligthing, and even masking specified areas.
Unlike traditional visual effects software (e.g., After Effects, Shake, Boujou, etc), the system is completely automatic and no manual labor is required from the user. The major limitation of the work is that it can currently handle only videos of static scenes (i.e., videos shot with a moving camera but containing no moving objects in the scene). Efforts are being made to lift this restriction in future work.
Pretty amazing results, especially if the software works as effortlessly as described. Unfortunately rendering times currently required are even more impressive – taking a Linux server farm 5 minutes to churn out a single frame! Developers say this will improve after forthcoming optimizations.