基于多窗体的改进视差图算法及其应用

An Improved Disparity Image Algorithm and Application Based on Multi-window

针对传统多窗视差图算法和大窗多窗视差图算法的误匹配率高、易受干扰和匹配速度慢等缺点,采用大型聚合窗固定和优化匹配代价计算方法的方式,提出了一种基于多窗体的改进视差图算法。该算法采用图像边缘提取和固定聚合窗相结合的方法,减少了匹配搜索空间;在匹配代价计算过程中,采用计算机缓存技术,提高了算法匹配速度和精度。试验表明相对于传统多窗算法,改进算法减小了图像的误匹配率和时间度,对于噪声以及复杂背景图像有较好的匹配结果、鲁棒性好,可应用于工作环境背景复杂、精度要求较高的工业和立体视觉领域。

The traditional multi-window algorithm and large window multi-window algorithm has high error rate, low robustness and low speed. In order to solute this disadvantages, a new algorithm of fixed large windows and optimization algorithm was proposed. The algorithm use edge extraction, fixed large window processes to reduce the algorithm search space, greatly improve the matching speed and accuracy. During computing matching cost, use the cache technology to increase matching speed. Experimental results show that the algorithm has lower mistakenly matching rate and higher efficiency. This algorithm can effectively filter noise and image edges effect and it has a practical utilization in a complex background and highly accuracy requirements area, such as industry application, computer version fields.