基於稀疏點視差估測與深度傳遞技術以建立雙眼內視鏡影像之全域深度圖

Using Sparse-Point Disparity Estimation and Depth Propagation to Construct Dense Depth Map for Stereo Endoscopic Images

黃希虹(HSI-HUNG HUANG)

指導教授 ： 賴文能

國立中正大學/工學院/電機工程學系/碩士(2016年)

摘要

微創手術是近年來非常受到重視的一項外科手術醫學技術，相比於產生大傷口的開放式手術，微創手術可以減少手術產生的傷口，也因此，微創手術已經逐漸成為外科手術主要發展的方向。然而如果單只提供 2D 畫面資訊，即使是技術非常熟練的醫師，微創手術難免也是會有些許風險，如同只讓醫師使用單隻眼睛在執行手術一般。如何提供準確的深度資訊，使得操刀醫師在執行手術的時候，有身歷其境的效果，從而獲得較佳的手術精度與較少的手術時間，是本篇論文應用的主要目標。
在本論文中，我們使用雙眼內視鏡攝影機所拍攝的影像序列，進行深度資訊的計算，進而能夠在多視角立體顯示器上呈現出真實的3D視覺效果。本論文所提出的深度資訊獲取是先透過獲取稀疏特徵點的深度資訊，再透過深度傳遞技術以獲得完整的深度影像。首先，我們透過SIFT特徵偵測與Canny邊緣偵測獲取稀疏點的位置，再透過左右影像間相似度匹配的計算獲得該些稀疏點的視差資訊，最後再透過雙邊濾波內差方法及空間傳遞優化的動作以獲取完整的深度影像。我們的稀疏點視差傳遞方法與傳統每一個像素都執行立體匹配方式相比，在同一視差平面可以達到較平滑的效果，所耗的時間也相對較短。

關鍵字

深度傳遞 ； 稀疏點 ； 立體匹配 ； 微創手術

並列摘要

In recent years, minimally invasive surgery has become very important in medical technology. As compared to a large open wound surgery, minimally invasive surgery can reduce the wound by surgery, therefore, minimally invasive surgery has gradually become a major medical surgery. However, if only have a single 2D image information, the physicians like with a single eye in the implementation of general surgery. Even if the skillful physician, minimally invasive surgery still have some risk. How to provide accurate 3D information is the main goal of this paper.
In this paper, we use the images within the sequence stereo endoscopic camera captured. We calculate the depth of information, and show real 3D vision in 3D stereoscopic display device. We use sparse point to get the sparse depth map, and then obtain dense depth map by bilateral filtering interpolated and refinement. First, we detect the SIFT and canny edge detector to obtain the location of the sparse point, and then calculating the similarity between the left and right images to get the sparse disparity of information. Finally, through bilateral filtering interpolated method and refinement of spatial propagation to obtain dense depth map. Our sparse disparity propagation method compared to a traditional stereo matching approach, can achieve a smoother effect in same disparity plane and the consumption of time is relatively less.

並列關鍵字

Depth propagation ； Sparse point ； Stereo matching ； Minimally Invasive Surgery

參考文獻

[2] Y. Boykov, O. Veksler and R. Zabih, “Fast Approximate Energy Minimization via Graph Cuts,” IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol. 23, No. 11, pp. 1222-1239, Nov. 2001.

[3] V. Kolmogorov and R. Zabih, “Computing Visual Correspondence with Occlusions via Graph Cuts,” Proc. of Int’l Conf. on Computer Vision, Vol. II, pp. 508-515, 2001.

[4] L. Hong and G. Chen. “Segment-based stereo matching using graph cuts,” Proc. of IEEE Conf. on Computer Vision and Pattern Recognition, vol. I, pp. 74-81, 2004.

[5] J. Sun, N.-N. Zheng and H.-Y. Shum, “Stereo Matching Using Belief Propagation,” IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 25, No. 7, pp. 787-800, July 2003.

[6] Q. Yang, L. Wang, R. Yang, H. Stewenius and D. Nister, “Stereo Matching with Color-Weighted Correlation, Hierarchical Belief Propagation and Occlusion Handling,” Proc. of IEEE Conf. on Computer Vision and Pattern Recognition, June 2006.