Film material may contain a variety of defects (e.g. flicker, noise, dust, instability). The defects which occur in almost any archived film are dust, dirt and blotches. Reliable detection of these single-frame defects is necessary for subsequent high quality defect removal.
The project Dust-NG develops novel algorithms and software tools for reliable detection and removal of single-frame defects like dust, dirt, blotches and hairs. The algorithms developed are optimized to process film scanned in high resolution (2K, 4K and above) in a fast and automated way.
Special attention is given on avoiding the introduction of any artifacts due to the restoration process. For this a variety of methods is introduced, e.g. the statistical analysis of the motion fields between the frames and appropriate mechanisms for handling strong film grain. All the applied methods result in a reliable solution for the detection and removal of single frame defects in an extremely automated way with a minimum of operator parameterization necessary.
The picture on the right shows the result image where single frame defects (dust, dirt, blotches, hairs?) have been restored.