Content generators, such as video producers and advertisers, can provide their content to TV networks and other publishers for display. Automatically determining if the provided content appeared on a specific TV network can be difficult. Matching captured frames from the video streamed by the TV networks to frames from the provided content can result in a low match rate. The match rate can be low because the TV networks often apply various image transformations, such as scalings, croppings, color transformations, and overlays such as logos, tickers and other on-screen graphics that can obscure the content. When the transformations are applied, the frame images ultimately displayed by the publisher don’t match the original frame images. The present paper discusses a system and method for determining, on a channel by channel basis, transformations and overlays that are applied by TV networks.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Volovich, Yaroslav; Oztaskent, Ant; and Rowe, Simon, "INCREASING AD DETECTION RELIABILITY BY LEARNING PER-CHANNEL TEMPLATES AND TRANSFORMATION", Technical Disclosure Commons, (January 13, 2017)