TV commercials are interesting in many respects: advertisers and psychologists are interested in their influence on human purchasing habits, while parents might be interested in shielding their children from their influence. In this paper, two methods for detecting and extracting commercials in digital videos are described. The first method is based on statistics of measurable features and enables the detection of commercial blocks within TV broadcasts. The second method performs detection and recognition of known commercials with high accuracy. Finally, we show how both approaches can be combined into a self-learning system. Our experimental results underline the practicality of the methods.