Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/84793
Title: | 基於積分成像式3D光場顯示器之設計與分析 Design and Analysis of Integral Imaging based 3D Light-Field Display |
Authors: | 許凱翔 Kai-Siang Hsu |
Advisor: | 林晃巖 Hoang-Yan Lin |
Keyword: | 光場顯示器,3D,積分成像,Elemental Image,深度,視覺輻輳調節衝突,元宇宙, Light-Field Display,3D,Integral Imaging,Elemental Image,Depth,VAC,Metaverse, |
Publication Year : | 2022 |
Degree: | 碩士 |
Abstract: | 隨著科技的進步,人們對於3D顯示的需求越來越強烈。從利用偏振眼鏡讓左右眼看見其對應影像,腦中產生立體視覺的穿戴式立體顯示(Stereoscopic),進化到利用雙眼視差所產生立體視覺的裸眼立體顯示(Autostereoscopic),最後演化成能在空間中重建光線資訊的光場顯示(Light-Field Display),都是人們對於3D顯示科技的追求。
全光函數包含了七個維度,是光場的數學模型,用來表示空間中任意一道光束。光場科技主要是由採集、顯示與資料處理全光函數組成。相較於傳統3D顯示技術,由於光場顯示技術能在空間中不同深度平面重建不同影像,能達成聚焦模糊,可以解決目前3D顯示技術中最主要的瓶頸-視覺輻輳調節衝突(Vergence-Accommodation Conflict, VAC)。 為了實現只需二維影像與深度圖生成EIs,本論文提出Computer Generated Elemental Image (CGEI) 演算法。與現有的演算法相比差異為增加權重函數與建立三維的離散相對座標,並使用蒙地卡羅光線追跡法進行光場重建影像的模擬與分析。增加權重函數可針對面板出光分佈找出EIs最佳解;建立三維的離散相對座標可自定義重建影像像素、透鏡陣列之各透鏡中心、面板像素的座標,能有更好的自訂性,未來對實際架設、曲面積分成像式光場顯示器、透明積分成像式光場顯示器皆有較高的相容性;使用蒙地卡羅光線追跡法進行光場重建影像的模擬與分析,在重建平面結果部份可提升約10dB的PSNR表現,驗證本系統可達成在空間中重建光線資訊,在模擬結果部份也有更完整的分析與討論。最終本論文實現不需要光場相機且避免視覺輻輳調節衝突的裸眼3D積分成像式光場顯示系統模擬。 With the advancement of technology, people's demand for 3D display is getting stronger and stronger. From the wearable stereoscopic display that uses polarized glasses to allow the left and right eyes to see the corresponding image and the following autostereoscopic display that uses binocular parallax to generate stereoscopic vision, 3D display technology finally evolved into Light-Field Display which reconstructs spacial light information. The plenoptic function contains seven dimensions, and it is a mathematical model of Light-Field, which is used to represent any light ray in space. Light-Field technology is mainly composed of acquisition, display and data processing of plenoptic functions. Because the Light-Field Display technology can reconstruct different images in different depth planes in space, it can achieve focus blur and solve the main bottleneck in the current 3D display technology, vergence-accommodation conflict (VAC), which is huge advantage comparing to the traditional 3D display technology. In order to generate EIs with only a 2D image and a depthmap, this paper proposes the Computer Generated Elemental Image (CGEI) algorithm. Compared with the existing algorithms, the differences are that Weight functions are added into it, the 3D discrete relative coordinates are established, and the Monte Carlo ray tracing method is used to simulate and analyze the Light-Field reconstruction image. Adding Weight functions into the algorithm can help us to find the best solution of EIs from the light distribution of the panel. Establishing 3D discrete relative coordinates has a contribution to the customization of the coordinates of the reconstructed image pixels, the center of each lens of the lens array, and the panel pixels. In the future, the proposed algorithm has high compatibility with the actual experimental, curved InIm Light-Field Display, and transparent InIm Light- Field Display. Using the Monte Carlo ray tracing method to simulate and analyze the Light-Field reconstruction image can improve the PSNR performance by about 10dB in the results, which verifies that the system can successfully reconstruct the spacial light information. Moreover, the analysis and discussion in the simulation results are also more complete. Finally, this paper realizes a simulation of naked-eye 3D Integral Imaging based Light-Field Display system that does not require a Light-Field Camera and avoids the VAC in the mean time. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/84793 |
DOI: | 10.6342/NTU202203008 |
Fulltext Rights: | 同意授權(限校園內公開) |
metadata.dc.date.embargo-lift: | 2027-08-31 |
Appears in Collections: | 光電工程學研究所 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-110-2.pdf Restricted Access | 10.83 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.