一种改进的多尺度多特征立体匹配算法
Improved Stereo Matching Algorithm Based on Multi-Scale and Multi-Feature
摘要: 在双目立体视觉领域中，立体匹配是其重要研究方向。为了针对部分立体匹配算法在弱纹理区域有较高的误匹配率的问题，本文提出一种基于多尺度多特征的立体匹配算法。将STAD、梯度与改进后的Census代价融合作为代价计算方法，代价聚合阶段，以引导图滤波算法为核心，利用跨尺度的思想将不同尺度的代价立方体进行融合，其中对于不同尺度的代价立方体设置了不同的代价聚合参数。对于视差图结果的一些错误，采用了多种视差后处理的方法。实验结果表明了该算法在弱纹理区域的准确性，对Middlebury3.0测试平台上标准图像对的实验结果表明，该算法在多组弱纹理图像上的平均误匹配率为8.16%，较传统的SGM等算法有更高的匹配精度。
Abstract: In the field of binocular stereo vision, stereo matching is an important research direction. In order to solve the problem that some stereo matching algorithms have high error matching rate in the weak texture region, this paper proposes a stereo matching algorithm based on multi-scale and multi feature. The STAD, gradient and improved Census cost fusion are used as the cost computing method. In the cost aggregation stage, take the guided filtering algorithm as the core. The cost cubes of different scales are fused using the idea of cross scales, and different cost aggregation parameters are set for the cost cubes of different scales. For some errors in disparity map results, a variety of methods of disparity post-processing are used. The experimental results show the accuracy of the algorithm in the weak texture area. The experimental results of standard image pairs on the Mid-dlebury 3.0 test platform show that the average mismatch rate of the algorithm in multiple groups of weak texture images is 8.16%, which has higher matching accuracy than the traditional SGM and other algorithms.
文章引用：岑亮, 秦襄培, 李祥鹏, 熊浩. 一种改进的多尺度多特征立体匹配算法[J]. 计算机科学与应用, 2023, 13(2): 259-269. https://doi.org/10.12677/CSA.2023.132026

参考文献

[1] Liu, C., Yuen, J. and Torralba, A. (2011) SIFT Flow: Dense Correspondence across Scenes and Its Applications. IEEE Transactions on Pattern Analysis and Machine Intelligence, 33, 978-994.
[2] Rhemann, C., Hosni, A., Bleyer, M., Rother, C. and Gelautz, M. (2011) Fast Cost-Volume Filtering for Visual Correspondence and beyond. CVPR 2011, Colorado Springs, 20-25 June 2011, 3017-3024. [Google Scholar] [CrossRef] 
[3] Yang, Q., Engels, C. and Akbarzadeh, A. (2008) Near Re-al-Time Stereo for Weakly-Textured Scenes. Proceedings of the British Machine Conference, 1-4 September 2008, 80-87. [Google Scholar] [CrossRef] 
[4] He, K., Jian, S. and Tang, X. (2010) Guided Image Filtering. Springer, Berlin. [Google Scholar] [CrossRef] 
[5] Scharstein, D., Szeliski, R. and Zabih, R. (2002) A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms. International Journal of Computer Vision, 47, 7-42.
[6] Mei, X., Sun, X., Dong, W., Wang, H. and Zhang, X. (2013) Segment-Tree Based Cost Aggregation for Stereo Matching. 11th European Conference on Computer Vision, Heraklion, 5-11 September 2010, 1-14. [Google Scholar] [CrossRef] 
[7] Yoon, K.-J. and Kweon, I.S. (2006) Adaptive Support-Weight Ap-proach for Correspondence Search. IEEE Transactions on Pattern Analysis and Machine Intelligence, 28, 650-656.
[8] Liu, H., Zhang, H.L., et al. (2021) Stereo Matching Algorithm Based on Two-Phase Adaptive Optimiza-tion of AD-Census and Gradient Fusion. 2021 IEEE International Conference on Real-time Computing and Robotics (RCAR), Xining, 15-19 July 2021, 726-731. [Google Scholar] [CrossRef] 
[9] Tomasi, C. and Manduchi, R. (1998) Bilateral Filtering for Gray and Color Images. 6th International Conference on Computer Vi-sion, Bombay, 7 January 1998, 839-846.
[10] Tan, P. and Monasse, P. (2014) Stereo Disparity through Cost Aggrega-tion with Guided Filter. Image Processing on Line, 4, 252-275. [Google Scholar] [CrossRef] 
[11] Hosni, A., Rhemann, C., Bleyer, M., et al. (2013) Fast Cost-Volume Filtering for Visual Correspondence and beyond. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35, 504-511. [Google Scholar] [CrossRef] 
[12] Kendall, A., Martirosyan, H., Dasgupta, S., et al. (2017) End-to-End Learning of Geometry and Context for Deep Stereo-Regression. IEEE International Conference on Comput-er Vision, Venice, 22-29 October 2017, 66-75. [Google Scholar] [CrossRef] 
[13] Chang, J.R. and Chen, Y.S. (2018) Pyramid Stereo Matching Network. IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, 18-23 June 2018, 5410-5418. [Google Scholar] [CrossRef] 
[14] Zhang, F., Prisacariu, V., Yang, R., et al. (2019) GA-Net: Guided Aggregation Net for End-to-End Stereo Matching. IEEE Conference on Computer Vision and Pattern Recognition, Long Beach, 15-20 June 2019, 185-194. [Google Scholar] [CrossRef] 
[15] Tombari, F., Mattoccia, S. and Stefano, L.D. (2009) Full-Search-Equivalent Pattern Matching with Incremental Dissimilarity Approximations. IEEE Transactions on Pattern Analysis and Machine Intelligence, 31, 129-141. [Google Scholar] [CrossRef] 
[16] Zabih, R. and Woodfill, J.I. (1994) Non-Parametric Local Transforms for Computing Visual Correspondence. European Conference on Computer Vision, Stockholm, 2-6 May 1994, 151-158. [Google Scholar] [CrossRef] 
[17] Hornung, A. and Kobbelt, L. (2006) Robust and Efficient Pho-to-Consistency Estimation for Volumetric 3D Reconstruction. European Conference on Computer Vision, Graz, 7-13 May 2006, 179-190. [Google Scholar] [CrossRef] 
[18] Zhang, K., Fang, Y., Min, D., et al. (2014) Cross-Scale Cost Aggrega-tion for Stereo Matching. IEEE Transactions on Circuits and Systems for Video Technology, 27, 965-976.
[19] Liu, H., Zhang, H.L., et al. (2021) Stereo Matching Algorithm Based on Two-Phase Adaptive Optimization of AD-Census and Gradient Fusion. 2021 IEEE International Conference on Real-time Computing and Robotics (RCAR), Xining, 15-19 July 2021, 726-731. [Google Scholar] [CrossRef] 
[20] Hirschmuller, H. (2005) Accurate and Efficient Stereo Processing by Semi-Global Matching and Mutual Information. IEEE Computer Society Conference on Computer Vision & Pattern Recognition, San Diego, 20-25 June 2005, 807-814.
[21] Scharstein, D., Hirschmuller, H., Kitajima, Y., et al. (2014) High-Resolution Stereo Datasets with Subpixel-Accurate Ground Truth. German Confer-ence on Pattern Recognition, Münster, 2-5 September 2014, 31-42. [Google Scholar] [CrossRef]