在大尺寸全像投影顯示器中,亞像素渲染及多屏拼接校正扮演著至關重要的角色。本論文中,我們提出一個基於影像辨識的全自動校正方法來進行這些校正程序。其中,本方法所提出之校正圖案用來便於校正重要之校正參數,包含柱狀透鏡之傾斜角度、寬度及中心位置,以及相鄰螢幕之間的間隙誤差。此論文所提出之方法會由一系列於市售全像投影螢幕進行之實驗來驗證,並且可以由這些實驗來看出我們所提出之方法對相機之位置、角度誤差,以及影像雜訊有著相當高的容忍度。最後,我們使用本論文提出之方法建構出一個54吋的大型全像投影顯示器,其可應用於許多場域,如影視娛樂及腹腔手術。 關鍵字:裸眼螢幕、子像素渲染、自動校正、多螢幕拼接
The calibration of subpixel rendering and multi-screen splicing plays a crucial role in the quality of large-scale autostereoscopic hologram display. In this thesis, an image-based method is proposed to perform such calibration procedures automatically. Particularly, specific patterns are designed to facilitate the calibration of critical parameters, including the slope, pitch, and center position of the lenticular lens sheet, and the misalignment gap distances between adjacent monitors. The proposed method is validated via a series of experiments performed on a commercially available 3D hologram display. Our method demonstrates a high tolerance for camera misalignment and image noises. Last but not least, a 54-inch hologram display is implemented using the proposed techniques. The developed prototype can be used in many applications such as entertainment and teleoperation. Keywords: Autostereoscopic display, Subpixel rendering, Automatic calibration, Multi-screen splicing