應用於三維立體電視之深度影像重建系統晶片設計與實現

Chip Design and Implementation of Depth Map Reconstruction for 3DTV System

邱鐘慶(Jung-Ching Chiou)

指導教授 ： 范育成

國立臺北科技大學/電資學院/電腦與通訊研究所/碩士(2010年)

摘要

目前的3D立體顯示技術主要是以眼鏡式3D技術為主流，但是配戴眼鏡畢竟比較不符合人性的需求，以長遠來說，裸眼式的顯示技術將會是未來3D立體電視發展的主流，而裸眼式3D技術需要大量的多視角影像，要有效壓縮這些多視角影像卻是一大挑戰，重建深度影像就是目前最有效的壓縮方式之一。
本論文採用三台平行式擺放的攝影機系統來重建深度影像，首先，利用階層式視差估計分別找出左右影像對中間影像的視差向量，因為利用較大的區塊在匹配搜尋時可以有較好的穩定性，但是較小的區塊確有較高的精準度，而階層式視差估計可以同時保有這些優點，以減少在區塊匹配時所產生的誤差以及遮蔽效應所造成的影響。接著對所產生出來的兩組視差向量做一致性檢查，偵測出匹配錯誤的地方並且填補合理的視差向量，此作法不但可以找出錯誤的視差向量並做修補，還可以使物件的輪廓在轉換成深度影像後更加精確。最後，我們採用三角量測演算法將視差向量轉換成深度值，即可重建高品質的深度影像。
本論文最重要的目標為：第一是希望能利用階層式視差估計來兼顧區塊匹配的穩定性及精確度，以產生精細的深度效果；第二是希望能利用「視差修補系統」偵測出匹配錯誤的視差向量，並且做合理的修正，減少深度影像破損的現象，以提高深度影像的品質；第三是希望能利用一張彩色影像以及其所對應的深度影像，來取代其他多視角的影像，進而有效地壓縮資料量。

關鍵字

視差影像 ； 深度影像 ； 視差估計 ； 區塊匹配

並列摘要

Glasses based 3D displays are familiar 3D product and technologies. However, human is unquestionably less receptive to wear glasses when we watch 3D video. In the long run, auto-stereoscopic display is foreseen to be the main 3D display technologies in the future. However, auto-stereoscopic display requires large amount data of multi-view images. To effectively compress and reduce multi-view images is a challenge. Reconstructing depth map is one of the more effective solutions to overcome this challenge.
This thesis introduces three-parallel-camera arrangement system to reconstruct depth map. At first, we adopt hierarchical disparity estimation to locate disparity vector of middle image using right view and left view information. This approach was taken during block matching. Larger block provides matching stability and smaller block supplies matching accuracy. Hierarchical disparity estimation in this case would ensure benefit to reduce errors during block matching and occlusion. Then, consistency check would detect mismatch block and fill in reasonable disparity vector. This approach would not only identify and repair errors for disparity vector but also produce clearer outline of images after transforming into depth map. Lastly, disparity vector is converted into depth value using triangulation algorithm to reconstruct high quality depth map.
The main goals of this thesis are to create fine depth image with hierarchical disparity estimation and to modify disparity increasing quality of depth map. This would ensure both stability and accuracy during block matching. Besides, disparity modification would reduce errors and provide reasonable correction for mismatching block and disparity vector. The presented scheme provides a color image and accuracy depth map for 3D video system. The proposed solution solves related problems and give a well solution in this thesis.

並列關鍵字

Disparity Map ； Depth Map ； Disparity Estimation ； Block Matching

參考文獻

[3] P. Kauff , N. Atzpadin , C. Fehn , M. Muller , 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.

[4] 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.

[5] C. T. E. R. Hewage, S. Worrall, S. Dogan and A. M. Kondoz, “A Novel Frame Concealment Method for Depth Maps Using Corresponding Colour Motion Vectors,” 3DTV Conference: The True Vision - Capture, Transmission and Display of 3D Video, pp. 149-152, May 2008.

[6] B. Kamolrat, W. A. C. Fernando, and M. Mrak, “3D Motion Estimation for Depth Information Compression in 3D-TV Applications,” Electronic Letters, vol. 44 no. 21, pp. 1244-1245, Oct. 2008.

[7] B. Kamolrat, W. A. C. Fernando, M. Mrak, and A. Kondoz,“3D Motion Estimation for Depth Image Coding in 3D Video Coding,” IEEE Trans. on Consumer Electronics, vol. 55, pp. 824-830, May 2009.