Soccer video processing for the detection of advertisement billboards

Billboards are placed on the sides of a soccer field for advertisement during match telecast. Unlike regular commercials, which are introduced during a break, on-field billboards appear on the TV screen at uncertain time instances, in different sizes, and also for different durations. Automated processing of soccer telecasts for detection and analysis of such billboards can provide important information on the effectiveness of this mode of advertising. We propose a method in which shot boundaries are first identified and the type of each shot is determined. Frames within each shot are then segmented to locate possible regions of interests (ROIs) – locations in a frame where billboards are potentially present. Finally, we use a combination of local and global features for detecting individual billboards by matching with a set of given templates.