Perceptual Color Film Restoration

Abstract

The cinematographic archives represent an important part of our collective memory. In the 1950s, monopack color film became the standard on which millions of cinematographic works were recorded. A couple of decades later, it turned out that this process was chemically unstable, causing the fading of whole film stocks with time. Since the bleaching phenomenon is irreversible, photochemical restoration of faded prints is not possible, hence the incontestability of digital color restoration. Usually, a bleached color release print is the only available record of a film, and no reference color is available; thus, color and dynamic range digital restoration is dependent on historical researches and on the skill of trained technicians who are able to control the restoration parameters. This can lead to a long and frustrating restoration process. For this reason, a restoration tool is a balance between a large number of complex restoration functions used to obtain accuracy in the result and a limit on the number of these functions to maintain simplicity in their use. As an alternative solution, we start from the robust capabilities of the human vision system (HVS) to propose a tool to filter damaged frames in a quasi-unsupervised way. In fact, film color cast, caused by aging, can be considered as generic chromatic noise, and thus a spatial color synthesis algorithm can be suitable for restoring it. Moreover, a method inspired by the HVS behavior does not need any a priori information about the color cast and its magnitude. Several tests have been performed with an algorithm called ACE (Automatic Colour Equalization). ACE is just one of the phases of the restoration pipeline, and it has been modified to meet the requirements of digital film restoration practice. The basic ACE computation autonomously extracts the visual content of the frame, correcting color cast if pres- ent and expanding its dynamic range. However, this behavior is not always a good restoring solution: There are cases in which the cast has to be maintained (e.g., underwater shots) or the dynamic range must not be expanded (e.g., sunset or night shots). To this aim, new functions have been added to preserve the natural histogram shape, adding new efficacy in the restoration process. Last, to complete the set, other functions have been added to obtain satisfactory results in cases where an input frame has been excessively corrupted. Examples are presented to discuss characteristics, advantages, and limits for the use of percep- tual models in digital movie color restoration.