Automatic Restoration Algorithms for 35mm Film

Peter Schallauer
Institute of Information Systems & Information Management
A-8010 Graz, Austria

Axel Pinz
Institute for Computer Graphics and Vision
Techische Universität Graz

Wener Haas
Institute of Information Systems & Information Management
A-8010 Graz, Austria



Abstract

Classical film restoration is based on special copying machines to improve the quality of safekeeping copies. Only a small class of defects can be removed by such a process, because the unit of manipulation is always the physical film strip. With help of digital image-processing techniques, the restoration process can be adapted for each frame or even pixel and also for many kinds of defects.

This paper presents algorithms for automatic and semiautomatic defect detection and removal in digital image sequences. First, an overview about typical defects in the 35 mm film domain is given. Then robust, full automatic algorithms to detect dust, dirt, and image vibration in noisy and disturbed environments are shown. To make possible expert user interaction with intermediate analysis data, detection and removal of defects are separated into a two-step approach. Automatic algorithms for dust and dirt removal and for image-sequence stabilization are presented as well as a semiautomatic algorithm for flicker and mold reduction. Finally, the algorithms are demonstrated and evaluated using several short film sequences.

Keywords: Digital film restoration, global motion estimation, dust, dirt, jitters, mold, flicker, local color variation, global color variation, 35 mm film