Translucent-static TV logo recognition by SUSAN corner extracting and matching

TV logo is widely used in copyright protection, video surveillance, etc. Traditionally, TV logo is detected and recognized by pixel-wise template matching, which neither deals well with translucent-static TV logos, nor addresses the scaling problem well. This paper proposes a novel TV logo recognition algorithm, which is composed of three steps. Firstly, the real time TV logo region is detected by multiple frame accumulation. Then, the Smallest Univalue Segment Assimilating Nucleus (SUSAN) algorithm is presented to extract corners within the detected logo regions. Finally, corner matching is carried out to recognize the type of the test TV logo. The proposed method is validated by several TV channels in China. The results indicate that the method is effective to detect and identify the translucent-static TV logos, and are invariant to image scaling.