Detection of on-field advertisement billboards from soccer telecasts
Alok Watve, Shamik Sural · 2006
We propose a method for the identification of on-field billboards meant for advertisements in soccer telecasts. We use a three-step approach in which the first step is to identify shot boundaries and classify the frames as part of a long shot or a close shot. The regions that are expected to contain billboards are then segmented with the help of domain knowledge and a number of heuristics. Finally, we use a combination of local and global features to identify the particular billboards of each advertiser. For each detected billboard, its position and the percentage of visible area are recorded. The results can be conveniently used to determine the effectiveness of billboard advertising.