【作者】 杜娟； 徐晟；

【Author】 DU JUAN;XU SHENG;College of Automation Science and Engineering,South China University of Technology;

【机构】 华南理工大学自动化科学与工程学院；

【摘要】 针对传统动态规划立体匹配算法匹配精度不高,且生成的视差图带有明显条纹瑕疵的问题,提出了一种基于改进自适应窗和动态规划的立体匹配算法,引入了颜色信息作为自适应窗口增长的约束条件,能够快速确定自适应窗口的匹配代价,有效提高了图像边界处的匹配精度。在用动态规划算法求取视差图之前,先利用图像分割技术获得图像的边界信息,在边界处建立新的平滑项模型,允许在边界处有较大的视差跳变。实验结果表明,该算法能够减少总体的匹配错误率,并且视差图横向条纹瑕疵明显减少,在视差不连续处能够得到较为准确的视差。更多还原

【Abstract】 The traditional dynamic programming stereo matching algorithm has problems such as lower matching-rate and obvious horizontal streaks of disparity map.This paper puts forward a new stereo matching method based on Improved Adaptive window and Dynamic Programming.Color information is introduced as a constraint condition for the growing of adaptive window to quickly determine the matching cost and improve the matching precision at the image boundary effectively.The dynamic programming algorithm is applied to obtain the disparity map after boundary information of the image is obtained from image segmentation technique and a new smooth term is established at the boundary which allows the disparity jump.The experimental results show that this algorithm can reduce overall matching error rates and apparent defective strips and accurate disparity can be obtained in disparity discontinuity regions.更多还原