近幾年,三維虛擬實境是一個非常熱門的研究領域,三維的技術非常大量地應用於相關電玩產業,但是大部分的三維產品或者研究,都不是使用真實的圖像或者影片來製作,而常常仰賴動畫,因此我們提出一套方法,採用真實二維圖像來模擬類似於三維的感官。我們的目標是能夠像Google Map實景圖的設計一樣地呈現物件,我們旋轉物件,而物件根據旋轉的角度對應至正確的影像。 一般來說,我們如果可以使用多台的攝影機,同時從周圍的各個視角拍攝物件,那麼對物件做各個視角之間的切換,會相當的容易,因為一起拍攝的關係,我們切換視角的銜接點是可以直接取得的,但是這樣的場邊配置、準備工作其實相當繁雜,因此我們主要的想法是只使用一台攝影機,拍攝一個人多個視角的影片,然後我們觀察人物走路的規律,並且分析人物的動作,最後,推導出視角與視角之間切換關係的規則。
The 3D virtual reality is a very popular research area over the last decade. The 3D products are proposed in video game industries. But most of the 3D products or researches didn’t use the real images or videos to generate, so we proposed a method to adopt the real 2D images to render 3D-liked vision. Our goal is to show the object like Google Map, we can rotate the object we load and according to the angle to show the corresponding object. Generally, if we can use more than one camera, and shooting the object from the surrounding each viewpoint. Then, the convergence of motion of switching angle will be very easy. Because we can directly obtain the matching time. But such preparatory work is too expensive. The main idea is shooting a human at multi-angle, then we observe the regular pattern of human and analyze the motion of human. Finally, infer a rule for the relationship between the change of viewpoint.