Aspects of Algorithm Development for Systems with Online Detection and Recognition of Tv Commercials

The paper describes a series of research stages in the development process for an automated system, able to detect commercial brakes in TV signals, and recognize a basic set of logos form the TV commercials. First, the conceptual diagram of the system is presented. Theoretical aspects of the recognition process are presented, and there are based on image matching algorithms. The paper presents also the test bench implemented to verify the correctness of the algorithm. The test bench is based on EasyAccess environment from eVision. Using the described test bench it is possible to tune the main parameters of the algorithm and realize extensive tests for different conditions and images. Using this test bench made possible to realize over 600 matching experiments using 20 logos of the most important international brands and 30 images captured in real conditions using different TV tuners.