A fast, parallelized logo detection algorithm on graphics processing units

Automatic logo detection in videos is required in different areas including automatic videos classification, automatic advertisement recognition, logo elimination in rebroadcasts and automatic detection of unpermitted broadcasts. Logo detection algorithms are expected to be fast and detect all kinds of logos with the least possible number of frames especially when they are used for detecting the logos in live videos. In this paper a filter designed which reinforce or weaken the edges of each frame of video based on the likelihood of existence of logo in the certain area of the frame resulting in logos to be detected with less number of frames. Moreover, the structure of proposed algorithm in this research is designed in a parallel format which is suitable for implementation on Graphics Processing Units (GPUs). Experimental results prove the algorithm to be an efficacious solution for automatic logo detection problems.