基于自适应非极大值抑制的SIFT改进算法

Improved SIFT algorithm based on adaptive non-maximun suppression

针对图像配准中尺度不变特征变换(SIFT)算法解算速率慢的问题,提出了基于非极大值抑制的改进算法。该算法扩大了极值检测区域半径,对SIFT关键点进行筛选,实现了关键点的优化分布。还提出一种自适应确定检测区域半径的方法,来更精确地控制关键点的数目和分布。仿真试验结果表明,该算法能在一系列不同的图像变换下表现出稳定的配准结果,解算速率较标准SIFT算法提升显著。

SIFT algorithm is inefficient in the applications wich have high rate of data updateing. This paper presents a non-maximum suppression algorithm to accelerate SIFT algorithm. This method extends the SIFT detector, so achieves well distributed key points. The results we obtained in tests demonstrate that the algorithm achieves high data renew rate and stable matching performance.