Practical Perceptually Adaptive Approach to Video Logo Placement in TV Broadcasts Alexander Wong Department of Electrical and Computer Engineering University of Waterloo Waterloo, Ontario, Canada, N2L 3G1 Email: a28wong@engmail.uwaterloo.ca William Bishop Department of Electrical and Computer Engineering University of Waterloo Waterloo, Ontario, Canada, N2L 3G1 Email: wdbishop@uwaterloo.ca Abstract— A common technique used by broadcast stations for video copyright protection is video logo placement, where a logo indicating ownership is inserted into the video broadcast in a visible way. This paper presents a practical algorithm that performs adaptive video logo placement in TV broadcasts based on a perceptual model from both spatial and temporal perspectives. The algorithm adjusts the position and the trans- parency of the video logo based on content activity, inter-frame relationships, and contrast with the goal of reducing overall perceptual loss while maintaining logo visibility. The algorithm is computationally efficient and suitable for video logo placement in live broadcasts. Experimental results show that the adaptive video logo placement algorithm achieves a good balance between perceptual quality and logo clarity. I. I NTRODUCTION Digital video broadcasting and distribution has highlighted the importance of copyright protection for digital video con- tent. An effective method used for indicating the ownership of video and image content is digital watermarking where additional information is embedded into a video or image. For copyright protection, the embedded information indicates the ownership of the digital media. For example, a corporate logo could be used to indicate ownership. Digital video and image watermarking can typically be divided into two categories: (i) Invisible watermarking [1]–[6], and (ii) Visible watermarking [7]–[12]. An overview of video watermarking techniques can be found in [13]. Invisible watermarking embeds information into the video or image content such that is not visible to the human eye. The main advantage of invisible watermarks is that they have a negligible effect on perceptual quality. However, invisible watermarks require special hardware or software to detect the watermark. Furthermore, such techniques are typically less robust to content manipulation. Common uses of invisible watermarks include content authentication and traitor tracing. Visible watermarking embeds information into the video or image content such that it is visible to the human eye. Unlike invisible watermarking, this technique reduces perceptual qual- ity. Common uses of visible watermarks include ownership identification and marking to deter unauthorized use. Most video watermarking research has focused on invisible video watermarking for compressed video content. Visible water- marking is less often the subject of research. One common approach used by broadcasting stations for video copyright protection is video logo placement, a form of visible watermarking where the logo of the owner is placed into a video broadcast in a visible manner. In the case of TV broadcast stations, the logo of the TV station is inserted into TV broadcasts. Video logo placement is widely used for TV broadcasts for a number of reasons. First, many TV broadcasts are live broadcasts, where video is delivered in “real-time” to the audience. Examples of live broadcasts include sporting events, award shows, and talk shows. The simplicity of video logo placement allows a TV broadcasting station to embed the station’s logo into the live broadcast stream with little processing time. Furthermore, video logos provide a very clear and explicit indication of video content ownership without covering a large area of the video broadcast content. Currently, video logos are positioned in a fixed location in all video frames of a TV video broadcast. There are several problems associated with placing a video logo at a fixed location in all video frames of a video broadcast. First, it is easy for someone to remove the video logo with little loss in video quality by cropping all video frames at a fixed location. Another problem is that the fixed logo placement may obstruct important video content that occurs at the fixed location. This leads to a loss in perceptual quality. One approach to reducing the perceptual quality loss is the use of semi-transparent video logos, such that the underlying video broadcast content can be partially viewed. However, the use of logos at a fixed transparency level for all frames may lead to significantly reduced logo clarity in some situations. This paper attempts to address all of the forementioned issues using a perceptual model, with the primary focus on retaining perceptual quality in the video broadcast after the placement of the video logo. While other adaptive video and image watermarking techniques have utilized perceptual models [2], [4]–[6], [10], [12], they are fundamentally different than the proposed algorithm. First, several of the previously researched techniques have been designed for the purpose of invisible watermarking [2], [4]–[6]. Of the techniques designed for visible watermarking [10], [12], none of them consider the perceptual quality and security implications of watermarking from a temporal perspective, which is important for video content. The proposed algorithm takes both spatial and tem- 0840-7789/07/$25.00 ©2007 IEEE 796