三眼カメラからの自由視点画像合成における奥行き知覚タスクを用いた視点操作手法の検討 (A Comparison of Two Methods of Viewpoint Manipulation in a Depth Perceptual Task)
View synthesis is an image-based rendering technique that involves generating a view of a scene from an arbitrary viewpoint based on multiple images captured from fixed viewpoints. View synthesis has been used for applications such as free viewpoint television, in which a user is free to “look around” at the transmitted video feed, and thus to perceive it from a variety of angles. By applying this view synthesis technique, we are able to provide users with a manipulable viewpoint of the scene, allowing them to access a continuum of virtual viewpoints. One of the important questions for users in such cases is how to manipulate the viewpoint in order to facilitate the continuous “looking around” effect, in particular to enhance the ability of users to experience motion parallax. To that end, we conducted an experiment to compare two different methods of viewpoint manipulation. The task involved subjects looking down at a pair of crossed rods, whose region of intersection was covered with an occluding surface. The subjects’ task was to peruse the scene to whatever extent desired, and then decide which rod was on top, i.e. closer to the viewer. In doing this, subjects were presented with two methods of continuous viewpoint manipulation. One involved a hand-held device comprising a pen slider and a tablet, as shown in the photo, with which subjects could manipulate the viewpoint by dragging the pen slider around. The other method, also shown in the photo, used the Microsoft Kinect to track the subject’s head and present the corresponding viewpoint. Because the Kinect is somewhat jittery when tracking the centre of the head, we predict that use of the Kinect will be perceived to be more difficult to use. Conversely, our expectation is that participants will perform better with the pen slider, which will be perceived as easier than the head tracking method. On the other hand, the effort of continuously holding the pen slider may make that method seem more stressful in terms of physical demands than the head tracking method. In our presentation, we will show the results, both objective and subjective, of our experiment to evaluate the two different methods.
越智 惇也 (大阪大学) , Walmsley William (University of Toronto) , Milgram Paul (University of Toronto) , 間下 以大 (大阪大学) , 清川 清 (大阪大学) , 竹村 治雄 (大阪大学)