基於深度影像重建技術實現二維至三維轉換系統與晶片設計

Chip Design and Implementation of 2D-to-3D Image Conversion System Based on Depth Map Reconstruction

張立承(Li-Cheng Chang)

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(2013年)

摘要

近年來，三維立體電視蓬勃發展，雖然現今3D成像技術已經達到趨近於成熟的設計方案，但是卻無法廣泛運用的原因在於呈現3D畫面的多視角影像在傳輸時有龐大的資料量，且現今拍攝的影像多數仍以二維影像為主，造成3D的數位內容寥寥無幾，在種種困難的前提下，多媒體系統很難支援3D的影像資訊。因此為了解決這個問題ATTEST(Advanced Three-dimensional Television System Technologies)提出了使用原始影像配合深度影像，最後經由DIBR演算法(Depth-Image-Based Rendering)來重建虛擬視角，即可將二維影像轉換為三維影像提供使用者在3DTV系統上觀看立體影像，讓使用者能輕易地感受到立體電視所帶來的樂趣。
本論文提出一個嶄新的二維至三維影像轉換技術，主要透過二值化與連通物件標籤化的方式，將二維影像分成物件資訊與背景資訊，利用Simple Linear Iterative Clustering(SLIC)方式，將物件影像作群聚，配合漸層深度影像，將物件與背景分別填上初始深度，最後再使用影像的邊緣資訊去修正初始的深度資訊，即可重建相對應的深度影像。除此之外，透過硬體架構的設計，將自動化的深度影像重建演算法實現於數位晶片上，加速立體影像的生成。

關鍵字

二維至三維影像轉換 ； 深度影像 ； 立體影像

並列摘要

In recent years, the field of Three-dimensional Television (3DTV) has been widely explored and developed. Though the technique of 3D display seems to be well-developed, it still cannot be widely adapted since the great amount of transmission capacity while being adapted to show multi-view image of 3D images, and the images nowadays are mainly two-dimensional images. These factors mentioned above caused the results of few 3D images. Under such circumstances, the current multimedia system used in this field could hardly support the process of 3D images. In order to best meet the urge need of the field, ATTEST (Advanced Three-dimensional Television System Technologies) proposed that using the original image with the depth map through Depth-Image-Based Rendering (DIBR) to reconstruct the virtual view so that the 2D images could be transferred into 3D images to provide a workable technique which could be used in state-of-art industry of televisions.
The current study aims at offering a brand-new workable 2D-to-3D image conversion which uses the thresholding function and image labeling to divide 2D images into two types of information, object and background. Then, use the method of Simple Linear Iterative Clustering (SLIC) algorithm to make the images cluster. By gradient depth map, assign initial depth value to the object and the background separately. Finally, through adapting the edge information of the image to amend the initial depth value, the corresponding depth map could be reconstructed. Moreover, through the design of the hardware, the 2D-to-3D image conversion could be branded into chips to expedite the reconstruction of stereoscopic images.

並列關鍵字

2D-to-3D Image Conversion ； Depth Map ； Stereoscopic Image

參考文獻

[3] T. Okoshi, Three-dimensional Imaging Techniques, Academic Press, NY, 1976.

[5] Dr Nick Holliman, 3D Display Systems, University of Durham, 2006.

[6] Charles Wheatstone Mirror Stereoscope

[8] 李宗熹，基於移動物件偵測實現二維對三維視訊轉換技術與晶片設計，碩士論文，國立台北科技大學，台灣台北，2012.

[10] T. Peterka, R. L. Kooima, D. J. Sandin, A. Johnson, J. Leigh, and T. A. DeFanti, “Advances in the Dynallax Solid-State Dynamic Parallax Barrier Autostereoscopic Visualization Display System,” in IEEE Transaction on Visualization and Computer Graphics, vol. 14, no. 3, pp. 487-499, May-June. 2008.