透過您的圖書館登入
IP:3.19.31.73
  • 學位論文

具品質資訊的次像元精度之影像邊緣線輔助影像密匹配

Subpixel Accuracy Image Edge with Quality Indicator Aided to Dense Stereo Matching

指導教授 : 趙鍵哲

摘要


利用攝影測量產製三維場景或物件乃常規性任務,然而為達致完整及高品質三維重建功效,幾何邊緣特徵對場景或物形的描述最具關鍵性。若能透過影像處理之邊緣偵測獲致完整且精確之幾何邊緣特徵,對於穩固影像密匹配視差不連續處與非視差不連續處有決定性之助益。尤其針對視差不連續處的視差劇烈變化,單一像元的不確定性就容易造成顯著的視差差異,導致前方交會產生嚴重位置偏差,更精確的次像元精度邊緣偵測有其必要性,因此本研究方法利用Edge Drawing搭配動量方程式偵測次像元精度邊緣線後,依據次像元精度邊緣品質使用十字區塊法約制局部式匹配的視差不連續處得到良好視差初值後,再透過改良式的半全域式匹配法優化視差,最後透過左右像檢查與填補以及改良式中值濾波獲取最終精化視差成果。其中本研究先透過視差真值資料萃取真實視差不連續處,來模擬完整及具次像元精度之幾何邊緣,驗證此高品質幾何邊緣資訊確實有助於獲致密匹配於視差連續與不連續區域之真實視差。在實際影像部份,亦針對室內影像及戶外近景與航拍影像進行本演算法之匹配成果分析,依此歸結帶有品質資訊的幾何邊緣對於影像密匹配的助益。此外,本研究提出的品質資訊指標有助於釐清所偵測的影像幾何邊緣是否能有效支援密匹配保留場景幾何邊緣特徵,依據本工作的實驗例所擬定的品質資訊指標門檻能指出弱邊緣特徵區域,達到成果評估功效,據此,亦可導引其他方法加入強化這些區域的匹配成效。

並列摘要


The integrity of geometric edge is vital to the high-quality 3-D scene representation. Thus, under the task of automated as well as successful 3-D scene or object reconstruction by photogrammetric approaches, edge detection becomes an indispensable essential. Various edge detection techniques can only provide the locations of edge up to one pixel accuracy level, which may still incur significant intersection error when dealing with acute depth discontinuities. In this study, Edge Drawing and moment equation are employed to detect subpixel accuracy edges. Then the depth discontinuities based on the subpixel accuracy edges’ quality are applied as constraints in local matching method to initiate a good approximation of disparity, and followed by the modified semi-global matching method to optimize the disparity. The maximum quality gain of subpixel accuracy edge detection aided to dense stereo matching, especially in depth discontinuities, is realized by the image pairs with ground truth disparity. The experimental data consists of indoor images, outdoor close-range images and aerial images. The results of various tests prove that the subpixel accuracy edges with quality indictors can benefit dense stereo matching effectively. The quality indicators of detected edges proposed in this study provide the information about the degree of high quality geometric edges in supporting dense stereo matching around depth discontinuities, thus weakly edge-constrained region can be identified and complementary alternatives can be further integrated to enhance the matching results.

參考文獻


李欣錡,2016。結合邊緣特徵資訊之半全域式匹配法,國立臺灣大學土木工程學系碩士論文,台北,台灣。
Agustín, T.P., Karl, K., Miguel, A.F., and Daniel, S.C., 2013. Accurate subpixel edge location based on partial area effect. Image and Vision Computing, 31(1), pp. 72-90.
Boykov, Y. and Jolly, M, P., 2001. Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images. Eighth International Conference on Computer Vision (ICCV 2001), pp. 105–112.
Boykov, Y., Veksler, O., and Zabih, R., 2001. Fast approximate energy minimization via graph cuts. IEEE Transactions on Pattern Analysis and Machine Intelligence, 23(11), pp.1222–1239.
Canny, J., 1986. A computational approach to edge detection. IEEE Transactions on pattern analysis and machine intelligence, (6), pp. 679-698.

延伸閱讀