適用於高動態範圍影像之立體匹配技術

Stereo Matching Techniques for High Dynamic Range Image Pairs

高崇傑(Chung-Chieh Kao)

指導教授 ： 林惠勇

國立中正大學/工學院/電機工程學系/碩士(2014年)

摘要

本論文專注於研究適用於高動態範圍影像之立體匹配技術。論文藉由修改三個現有具代表性之立體匹配演算法，並建立高動態範圍影像來進行立體匹配。我們從實驗結果中探討這過程中會影響計算結果之原因與注意事項，接著針對高動態範圍影像進行位元平面切割發現僅需每頻道16位元即可儲存大部分高動態範圍影像。最後提出一個每頻道16位元的整數表示法來存高動態範圍影像資料，藉由這種儲存方式能讓更多現有之立體匹配演算法直接使用高動態範圍影像資訊來進行立體匹配之相關研究。

關鍵字

高動態範圍影像 ； 立體匹配

並列摘要

In this thesis we focus on stereo matching techniques for high dynamic range image pairs. We start by pointing out key points for generating HDR images suitable for achieving high-quality stereo matching results. Then 3 state-of-the-art stereo matching methods are modified and used to test the robustness of the proposed HDR stereo compared to conventional LDR stereo matching results. By performing HDR bit-plane slicing we found that we needed only around 16 bits per channel to store HDR images. At the end we propose a 16-bit unsigned integer format to store HDR pixel data, thus allowing many available stereo matching algorithms to be tested with HDR stereo pairs with only slight modifications.

並列關鍵字

Stereo Matching ； High Dynamic Range Imaging ； HDR

參考文獻

for avoiding obstacles during visual navigation,” in Robotics and Automation

(ICRA), 2012 IEEE International Conference on, pp. 4850–4855, May 2012.

[4] S. Hermann and R. Klette, “Iterative semi-global matching for robust driver

[5] S. Hrabar, P. Corke, and M. Bosse, “High dynamic range stereo vision for outdoor

[7] D. Scharstein and C. Pal, “Learning conditional random fields for stereo,” in