基於動作捕捉的虛擬場景合成系統

在商業攝影拍攝當中，經常搭建或租用實體場景，輔助消費者了解商品的使用情境，但實體的場景籌備過程耗時及昂貴。本論文將探討背景合成的技術、光學成像、消費等級的動態捕捉設備、電腦繪圖技術，整合出簡易操作且低價的合成系統(MocapVSCS)，達到商品實體背景的數位化。最終將系統所產生的影像，與現場拍攝的實景，投入量化與質性兩種評估方式進行評比，探討是否能夠達到無法辨別實景拍攝與合成影像的品質。最終本研究的系統MocapVSCS所產生的結果，在量化評估中取得了95.95%的高相似度，質性評估的部分也取得了40.56 % 欺騙指數，接近最高為50%的理想值，並且在MocapVSCS的協助下可以降低實際拍攝費用約46%與減少約38%的製作時間。

In commercial photography shooting, physical scenes are often built or rented to assist consumers in understanding the use situation of the product, but the physical scene setting up process is time-consuming and expensive. This thesis will provide a literature review of background synthesis technology, optical imaging, consumer-grade motion capture equipment, and computer graphics technology and integrate a low-cost and straightforward synthesis system (MocapVSCS) to digitize the product's physical background. Finally, the image produced by the system compared with the actual scene shot on-site, using two evaluation methods, quantitative and qualitative, to explore whether the difference between the actual scene shot and the synthesized image can be distinguished. In the end, results produced by the system MocapVSCS of this study achieved a high similarity of 95.95% in the quantitative evaluation, and the qualitative evaluation part also achieved a deception index of 40.56% , which is close to the ideal value of the highest 50%. with the help of MocapVSCS, the cost of shooting can be reduced by 46% and the production time can be reduced by 38%.