Image processing for restoration of old film sequences

The old film restoration involves the development of image processing technology. We focus on the correction of film misalignment and the removal of blotches. In the restoration, the film is first read with a scanner, but the scanned sequence often suffers from irregular spatial vibration due to inaccurate frame alignment. This artifact is perceived as a flicker. Reduction of this flicker needs accurate knowledge of relative locations of scanned frames. Global interframe deformation contains such information. However, interframe deformation is also caused by camera work. Hence, we develop a robust correction method for estimating interframe misalignment separate from camera work and compensating for it. After the correction of interframe misalignment and the separation of camera work, the blotch removal can be performed by a filtering method. We present a spatiotemporal nonlinear restoration filter whose response is controlled according to the local anisotropic brightness continuity of a corrupted sequence.