透過您的圖書館登入
IP:18.222.162.216
  • 學位論文

利用動態規劃方法重建影像深度之系統晶片實現

Chip Design and Implementation of Depth Map Reconstruction Using Dynamic Programming

指導教授 : 范育成
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


目前3D顯示技術中,可以大略分為裸視型與需配戴眼鏡的兩種分類。裸視型顯示技術具有先天的優勢,可以免去眼鏡的困擾,較為被普遍使用者接受,但裸眼式3D顯示需要大量視角的資訊,來減少遮蔽效應的產生。因此資料量的壓縮變成一項重要的議題。利用影像深度重建多視角影像,是目前多視角影像最有效的壓縮方法。 本論文使用兩台平行攝影機,尋找相對的視差向量,搜尋視差向量可分為區域性與全域性的搜尋,區域性的搜尋受區塊大小所限制,在控制區塊大小方面是最大的問題所在,而全域的搜尋通常會產生大量的計算量,但沒有區塊大小適應性的問題,我們使用左右平行兩張影像進行動態規劃的視差搜尋,利用影像的亮度值,以左右視角的水平掃描線,並以兩個化簡的矩陣紀錄各點累積的誤差值,同時與前一個可能相關的節點,搜尋全域最小誤差累積,得到最佳路徑,找出合理且稠密的視差向量。 本論文的目標,希望能利用動態規劃的方法,透過兩個平行攝影機來實現影像的視差向量搜尋,利用演算法的化簡與硬體平行架構與資料回授的數位晶片設計,來達到影像解析度640×480上視差搜尋演算法的即時處理,產生稠密且準確的深度圖。

並列摘要


3D displays technology become an important research field recently. 3D displays are divided into auto-stereoscopic and 3D glasses two categories roughly, there are many different 3D display principles have been gradually proposed. Auto-stereoscopic can be removed glasses problems require large amount data of multi-view images to reduce occlusion. So data compression becomes an important topic. Reconstructing depth map is an effective solution to solve the challenge. In this thesis, we use binocular stereo camera system to search relative disparity vector which classify local and global algorithm. In local disparity estimation, we need define the search block size. But decision of matching block size is a major problem. And the global disparity estimation doesn’t reach real-time processing because it makes huge computation. In order to solve the problems, a reduce algorithm is proposed to improve dynamic programming algorithm to reconstruct dense depth map. In this thesis, we perform chip design and implementation of depth map reconstruction using dynamic programming and reach real time 3D video processing.

參考文獻


[10] 林品志,應用於多視角影像顯示系統虛擬視角影像生成技術之演算法及硬體架構實作,碩士論文,國立台灣大學電子工程學研究所,台北,2010。
[5] P. Kauff , N. Atzpadin , C. Fehn , M. Müller , O. Schreer , A. Smolic and R. Tanger, “Depth Map Creation and Image-Based Rendering for Advanced 3DTV Services Providing Interoperability and Scalability,” Image Communication, vol. 22, no. 2, pp. 217-234, Feb. 2007.
[8] C. L. Zitnick and T. Kanade, “A Cooperative Algorithm for Stereo Matching and Occlusion Detection,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 22, no. 7, pp.675-684, Jul. 2000.
[11] 龔毓婷,裸眼式三維立體顯示發展系統,碩士論文,國立台北科技大學電腦與通訊研究所,台北,2010。
[12] D. Scharstein, R. Szeliski, and R. Zabih, “A Taxonomy and Evaluation of Dense Two-frame Stereo Correspondence Algorithms,” IEEE Workshop on Stereo and Multi-Baseline Vision, Kauai, USA, Dec. 9-10, 2001, pp. 131-140.

延伸閱讀