摘要
将小波变换运用到立体视觉的图像匹配中,选取合适的二进小波基,以二进小波变换的多尺度零交叉为匹配基元,较多地利用了图像的结构特征.在匹配过程中采用由粗到精的层次匹配策略和严格的匹配约束,减小了对应点搜索范围,并从局部和整体2个角度来解决歧义匹配的问题.实验结果证明,该方法具有较高的匹配精确度和较快的匹配速度,并且算法稳定.
Proposing an algorithm for stereo-matching, which utilizes the wavelet transform. The multiresolution zero-crossing of binary wavelet transform are used as matching primitives, the algorithm uses a coarse-to-fine matching strategy and strict constraints. Experimental results show that the algorithm is reliable and stable in operation, and has high accuracy.
出处
《装备指挥技术学院学报》
2000年第3期38-43,共6页
Journal of the Academy of Equipment Command & Technology
关键词
计算机视觉
立体匹配
小波变换
多尺度零交叉
computer vision
stereo-matching
wavelet transform
multiresolution zero-crossing