An adaptive anchor frame detection algorithm based on background detection for news video analysis

When analyzing news videos, finding an efficient way of extracting visual memes is very important. Videos might be very long and visual meme extraction itself is computationally expensive, so it is essential to make this process as efficient as possible. A way to do this is to eliminate as many key frames as possible even before extracting the visual memes. Since anchor person frames contribute little to the content of the news videos, we should remove these frames. This paper proposes an efficient and effective algorithm to detect anchor frames from videos, significantly improving the efficiency of visual meme extraction.