基于加速鲁棒特征的木材显微图像自动配准方法

针对显微镜观测视野狭小而难以采集到全局图像的问题,提出了一种基于加速鲁棒特征(SURF)的木材显微图像自动配准方法.首先使用SURF检测并描述兴趣点,通过最近邻匹配得到匹配点对后,用双向匹配和RANSAC算法剔除错误匹配.然后利用最小二乘法和匹配结果进行模型参数估计,最后通过插值获得配准图像.对阔叶材显微图像配准实验结果表明,该方法具有较好的鲁棒性,无论图像是否有旋转,都可以实现自动的配准.比起尺度不变特征转换(SWT),由于用SURF得到的兴趣点数量更少,运算速度更快,总的匹配速度提升了5倍左右,缩短了整个配准过程的时间,算法更具有实时性.

A new wood microscopic image registration approach based on speeded up robust features （SURF）

To solve the problem of narrow fields for microscopes, a new method of microscopic image registration based on speeded up robust features （SURF） was developed. First, feature points were extracted using SURF, and corresponding matching points were found using the nearest neighbor method. Wrong matches were eliminated using the Random Sample Consensus （RANSAC） algorithm and bilateral matching. Then, transformed parameters were estimated using least squares techniques and matching the results. Finally, through interpolation, the registered image was achieved. Registration results of multiple wood images showed that the algorithm was robust. Regardless of whether the image is rotated, automatic registration can be achieved. The accuracy of our method is comparable with SIFT （scale-invariant feature transform） image registration method while registration is about 5 times faster. Therefore, alzorithm is more real-time. Ch. 4 fig. 1 tab. 17 ref.]