基於內容之色彩加深度圖三維影像尺寸調整架構

Content-aware Color plus Depth Map 3D Image Resizing

周瑋慈(Wei-Cih Jhou)

指導教授 ： 吳家麟

國立臺灣大學/電機資訊學院/資訊網路與多媒體研究所/碩士(2011年)

https://doi.org/10.6342/NTU.2011.02611

摘要

最近，有許多研究意識到媒體尺寸調整的問題，由於可裸眼觀看3D影像的個人數位顯示器之蓬勃發展，媒體內容不只是要在一般顯示器上做調適，也必須在3D顯示器上做調適。與傳統的單視圖相比，顏色加深度圖(或視差圖)三維圖像能為我們提供更多有關立體觀感的資訊。因此，單視圖的傳統視覺注意力分析模型，難以直接應用於調整顏色加深度圖的三維圖像上，再加上對於觀看一個3D內容而言，其深度資訊的立體視覺舒適區域，針對不同3D顯示器裝置的螢幕大小也有所差異，如果不考慮立體視覺舒適區域，3D視覺體驗將是充滿壓力與不舒適的視覺感受。
在本論文中，我們首先提出一個自下而上的顯著值模型，該模型結合深度的資訊來模擬立體人類視覺系統，然後，將此顯著性模型與現有的縮放技術整合進而提高調整大小後的二維圖像之品質。最後，為了將不同大小的3D屏幕空間沿著Z軸方向壓縮，我們對深度資訊做一個非線性的映射技術，使產生的深度分布，落在觀看者的立體視覺舒適區域內。
本論文所提出的三維影像尺寸調整技術將人類視覺注意模型的立體感知效果(顯著性和舒適區)考慮在內，實驗結果顯示，調整大小後的二維圖像之品質將會被增進，因此，也同時會增進人類的3D視覺體驗。

關鍵字

尺寸調整 ； 視覺顯著性 ； 立體視覺舒適

並列摘要

Recently, there are numerous works focusing on content aware media resizing. Due to the amount of personal digital displays on which one can watch 3D images with naked eyes grows rapidly, media need to be adapted not only to regular display but also to 3D display. Compared to traditional mono-view photographs, color plus depth map (or disparity map) 3D images provide us with more information about the stereoscopic perception. Therefore, traditional visual attention model for mono-view photographs can hardly be applied to resize the color plus depth map 3D images directly. The stereo comfort zone of the depth information for viewing 3D contents varies with the size of 3D display devices. Without taking the comfort zone into consideration, the 3D viewing experience will be stressful to the stereo vision. In this thesis, we first propose a bottom-up saliency model, which incorporating with the depth information to simulate the stereoscopic vision of the human visual system. Then, the saliency model is integrated into an existing resizing technique to enhance the quality of the resized 2D image. Finally, to squeeze the various 3D screen spaces along the z direction, a nonlinear depth mapping technique is applied to make the resultant depth fall into the comfort zone. By taking the stereoscopic perception effects (saliency and comfort zone) of the human visual attention model into account, experimental results show that the quality of the resized 3D images will be improved which, in turn, will enhance our viewing experience for 3D images.

並列關鍵字

resizing ； visual saliency ； stereo visual comfort

參考文獻

[1] R. Achanta and S. Susstrunk. Saliency detection for content-aware image resizing. In IEEE International Conference on Image Processing (ICIP), 2009 16th, pages 1005 –1008, Nov. 2009.

[2] S. Avidan and A. Shamir. Seam carving for content-aware image resizing. ACM SIGGRAPH 2007 papers, 2007.

[3] C. Barnes, E. Shechtman, A. Finkelstein, and D. B. Goldman. Patchmatch: a randomized correspondence algorithm for structural image editing. ACM SIGGRAPH 2009 papers, pages 24:1–24:11, 2009.

[5] T. S. Cho, S. Avidan, and W. Freeman. The patch transform. IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(8):1489–1501, 2010.

[9] Y. Guo, F. Liu, J. Shi, Z.-H. Zhou, and M. Gleicher. Image retargeting using mesh parametrization. IEEE Transactions on Multimedia, 11(5):856 –867, Aug. 2009.