Video signals originating from video cameras have properties that differ from video originating from film-scanners. For advanced scan-rate conversion, it is important to distinguish between the two types and adapt the processing to the type. Incorrect processing in such scan-rate converters could lead to undesirable artefacts in the converted video signal. Therefore, a robust method is required to classify video signals, based on their origin. Since this method applied in consumer electronics, such as high-end televisions, a cheap implementation is also required. In this project, such detection mechanisms, better known as film detectors, are examined and compared.
Based on the available literature, an overview of the field of film detection is given. The Schutten-Riemens Film Detector (SRFD) is examined.
· A new difference metrics is proposed. This difference metric is insensitive to static vertical detail and has only a 8.5\% increase in cost, compared to the original SRFD.
· A hard cut shot boundary detection scheme is implemented and tested.
· A new film mode class is proposed: The hybrid mode. This mode describes mixed video sequences that have both the characteristics of video mode and film mode.
· A new source of mixed mode sequences has been identified. The encoding process in DV-cameras produces a motion pattern that can be detected as 2:2 pull-down film mode by the film detector.