隨著三維立體顯示技術(Three-Dimensional Stereoscopic Display Technology)的興起,眼鏡式立體顯示技術,其觀看的便利性與舒適度備受限制,裸眼式立體顯示技術(Auto-stereoscopic Display Technology)將成為發展的重點與目標。 本篇論文以裸眼式立體顯示技術為基礎,提出一套三維立體影像合成系統(Three-Dimensional Stereo Image Synthesis System ),由九個不同視角所拍攝出的影像作為輸入來源,經由所提出的子像素排列對映架構(Mapping Architecture),每個子像素經由精密與準確的計算,可以正確地指定每個像素的位置和次序,將二維影像轉換為多視角三維立體影像,再配合多視角裸眼式立體顯示器,即可呈現出具有立體感知的視覺效果;此外,並將此二維轉換三維的合成架構,透過系統發展平台驗證與實現其顯示流程,也就是利用現場可規劃邏輯陣列(Field Programmable Gate Array,FPGA)元件做為系統設計驗證的工具。 本篇論文所提出的對映架構,可以輸入九張不同視覺位置所拍攝出來的影像資訊,經由此演算架構,將二維影像轉換為多視角立體影像,在目前深度影像還未能準確的提供的情況下,完全不需要任何深度資訊的多視角合成方式,改善了深度資訊不足的缺點,節省傳統立體影像在製作上所耗費的時間。
The development of stereoscopic display technology has led to 3D related research. Auto-stereoscopic display technology provides convenience and comfortable vision and plays an important role of 3D display. This thesis develops a 3D synthesis and mapping technology based on lenticular auto-stereoscopic display. The system captures nine-view images (two-dimensional images) according to multi-view angle. Then, multi-rate signal processing is performed to scale the images’ size. Afterward, we establish a recursive-mapping rule based on sub-pixel arrangement and regulation of 3D display. Besides, the proposed scheme considers precision and accuracy of sub-pixel mapping and provides realistic vision. In order to achieve the presented algorithm, we adopt Xilinx FPGA development platform to perform the 3D image synthesis and mapping system. We propose a 3D image mapping architecture that adopts nine-view image data according to the multi-view image arrangement rule without depth map information in this thesis. The proposed scheme uses recursive-mapping technique and reduces hardware overhead to provide better solution for auto-stereoscopic display. The experimental results show this approach appearing realistic 3D vision and providing the efficiency solution of 3D synthesis research.