| 摘 要: 针对传统Census变换在立体匹配中存在的局限性,即不能充分利用邻域窗口信息且过于依赖中心点像素,以及图像在受到噪声干扰时窗口中心值的改变将导致全局匹配精度降低的问题,提出了一种基于改进Census变换的半全局(Semi-Global Matching,SGM)立体匹配算法。首先,融合了绝对差值之和(SAD)算法与Census变换算法,并引入自适应窗口和噪声容限参数,以增强邻域窗口信息的相关性、减少对窗口中心点的依赖,并提升代价计算的可靠性;其次,采用八路径代价聚合与赢者通吃(WTA)算法获得初始视差;最后,利用加权最小二乘(WLS)滤波及剔除小连通区等方法,对初始视差图进行优化,得到最终的视差图。实验结果表明,本文所提出的立体匹配算法在全区域的平均误匹配率仅为8.02%,有效降低了误匹配率,显著提高了抗噪干扰性能。 |
| 关键词: Census变换;立体匹配;SAD;自适应窗口;噪声容限参数 |
|
中图分类号: TP391
文献标识码: A
|
|
| SGM Stereo Matching Algorithm Based on the Fusion of SAD and CensusTransform |
|
XING Yue1, ZHU Jiezhong1, YANG Zaiqiang2, YAO Chengmin1, ZHANG Mingjie1
|
(1. School of Automation, Nanjing University of Information Science and Technology, Nanjing 210024, China;
2. School of Applied Meteorology, Nanjing University of Information Science and Technology, Nanjing 210024, China;)
764864509@qq.com; 1450466566@qq.com; 691071467@qq.com; 2467600365@qq.com; 2497167418@qq.com
|
| Abstract: The limitations of the traditional Census transform in stereo matching include its inability to fully utilize neighborhood window information and its excessive reliance on the center pixel values, as well as the problem that changes in the center value of the window under noise interference can lead to a decrease in global matching accuracy.In view of these problems, this paper proposes a Sem-i Global Matching (SGM) stereo matching algorithm based on an improved Census transform. Firstly, the Sum of Absolute Differences (SAD) algorithm and Census transform are fused, and adaptive windows and noise tolerance parameters are introduced to enhance the correlation of neighborhood window information, reduce dependence on the window center, and improve the reliability of cost computation. Secondly, an eigh-t path cost aggregation strategy and the Winne-r Takes-All (WTA) algorithm are employed to obtain the initial disparity. Finally, the initial disparity map is optimized using Weighted Least Squares (WLS) filtering and small connected region removal to produce the final disparity map. Experimental results demonstrate that the proposed algorithm achieves an average mismatch rate of only 8.02% across all regions, effectively reducing mismatch rates and significantly enhancing noise resistance. |
| Keywords: Census transform; stereo matching; SAD; adaptive window; noise tolerance parameter |
