Giorgio Trumpy, Andreas Wassmer, Rudolf Gschwind
Imaging & Media Lab – UNI BASEL
The main topic of the DUST_BW project relates to digital reconstruction and permanence of photographs (slide, negative) and motion-picture films by digital image processing. Several factors limit the permanence of photographic material: this project focuses on dust and scratches. The goal is the development of a scanning system that will efficiently detect, and subsequently remove, dust and scratches from any type of transparent photographic material (color as well as black-and-white, still images as well as moving images). The removal of these defects is performed through a “hardware assisted” method that uses two techniques: crossed polarization and dark field setup.
State of the Art Edit
Different techniques are used to remove defects on photographic material, each has its advantages but also its limitations.
Mechanical dust removal techniques utilize pressurized air combined with electrostatic discharge and/or applying the film on a dust removal roller with a sticky surface; this is a low-budget method but the efficiency is not very high, especially when dirt is sticking to the emulsion or film base.
Other methods use chemical substances, like the immersion of the film into a cleaning solvent, or the so called “WetGate” method; the main drawback of these methods is the fact that the chemicals used are environmentally and toxicologically hazardous, which makes movie restoration quite expensive.
Since several years, with the strong spreading of digital imaging technologies, digital methods for restoration are replacing analog techniques. A well-known method consists in the automatic removal of dust using the timeline. Dust has an impulse answer, i.e. appears only on one frame. The comparison of a movie frame with the frame before and after allows detecting dust. Nevertheless, if the image has a very fast motion, it is quite complicated to figure out if a specific element in a scene is a dust spot or just a fast moving object and this can entail false detections.
The application of filters on digital images (e.g. median, despeckle) can be quite effective but have the disadvantage that it cannot distinguish what is actually dust or noise. It will also obliterate fine detail in the image.
In the case of color photographs, however, “infrared cleaning” can be used. This method consists in collecting an additional infrared channel from the scan at the same time as the visible color channels; photographic color film is mostly transparent to infrared radiation (no matter what is the image it contains), conversely dust and scratches are not, so they are recorded in the IR channel. A major limitation of this technique is that it can only be used on dye-based films (colour and chromogenic black-and-white films); the image-forming silver particles in black-and-white film stocks, on the other hand, are opaque to infrared radiation, making this method inapplicable.
Moreover, this technique is patented and this entails extensive additional costs.
Another solution, known as Dirt & Scratch Correction Option (D/SCO), involves the utilization of an integrating sphere that provides a perfectly diffused illumination to the film. A portion of the interior surface of the integrating sphere is arranged to have a variable reflectivity; this characteristic allows balancing two different signals corresponding to the scattered and unscattered light. This feature can be used to electronically reduce the appearance of blemishes.
Other solutions have been developed for dust and scratch correction involving dark field illumination and crossed polarization. Nevertheless no documents have been found regarding the combination of these two techniques.
Description of the project Edit
Dust and scratches can be considered as two different types of perturbation of the parallelism of the film surfaces: a dust grain is additional material on the film surface and, on the contrary, in a scratch the film material is missing. In order to perform the defects detection, an image can be acquired with a special optical set-up specifically designed to localize exactly irregularities present in the surface of the film.
The size of developed silver grains in black-and-white emulsion is < 1μm, whereas dust particles and scratches are generally much bigger. By setting an angle between the illumination and the optical axis of the acquisition device (dark-field setup), it is possible to exclude the unscattered beam from the image. Therefore, since the silver particles and the defects scatter light to a different extent, by adopting a dark-field setup it is possible to obtain an image in which the signal related to the silver emulsion is considerably suppressed and the defects are emphasized.
If polarized light is used, depolarization of scattered light is observed and this phenomenon is known to depend on the nature and size of the scatterers. The rays scattered by dust particles and scratches are depolarized to a greater extent than rays scattered by developed silver grains. Therefore, the dark-field setup can be combined with the cross-polarization technique, thus obtaining an improvement of the defect detection: the analyzer blocks the scattered light from the silver grains and gives the possibility to have an image in which the signal of the silver emulsion is further suppressed.