Semi-transparent watermarks (logos) are a prominent feature of modern television programs. They are employed to label broadcasted programs, but not commercials, and therefore their absence can be exploited to filter out advertisements. A real-time template-based pixel matching algorithm for detecting known logos in video is presented. The logo detector achieves a high degree of precision (99.53%) and recall (97.23%) despite significant spatial variations inherent in the decompressed source video. When combined with other feature detectors, it forms the foundation of a live commercial detection system with applications in video archiving and embedded environments.
Citation:
Balint Seeber, Neil Yager, Adnan Amin, "Real-time Detection of Semi-transparent Watermarks in Decompressed Video," wacv, pp.49, Eighth IEEE Workshop on Applications of Computer Vision (WACV'07), 2007