| Summary: | 碩士 === 國立交通大學 === 多媒體工程研究所 === 101 === Spot-the-Difference is a game, in which players need to find the differences between a pair of similar images. The two images look almost the same at first glance but actually have some differences if one carefully observes them. Besides static images, Spot-the-Difference (STD) games can also be played on a pair of similar videos. We call this type of games dynamic STD games. Compared to static STD games, it is more difficult to design and make videos for dynamic STD games because they involve video editing. Hence, to the best of our knowledge, there have not been effective and systematical approaches for making videos for dynamic STD games.
In this thesis, we propose an approach that can generate videos for dynamic STD games at different levels of difficulties based on an input video recorded at a fixed camera view. According to the motion-Induced blindness theory, we define the attention level (ATL) and the awareness level (AWL) of motion segments by computing the video saliency map from the input video. Based on ATL and AWL, we can control the level of difficulty by choosing and moving appropriate motion segments as the differences for players to spot. In our experiments, we used an eye tracker to verify the necessity of each factor considered in our definition of ATL and AWL. We also conducted user studies to validate our results, which show that players do need more time to complete a STD game of greater difficulty.
|
|---|
