关键词: 特征向量, MODIFIED HAMMER算法, 参数特征, 超声图像配准

Abstract: Ultrasound images registration has attracted more and more attentions nowadays.Due to the speckle noise in ultrasound images，it is difficult to detect features via the calculation of their derivatives.A new parameter feature，which is obtained from the statistical properties of speckle noises，is utilized to replace the gradient amplitude.Firstly the leading points are extracted by feature vectors，which are constructed by the intensity，the Laplacian of the Gaussian and the new parameter feature.Then ultrasound images are registered based on the framework of MODIFIED HAMMER.Experimental results show that the novel method can effectively reduce computational time by cutting down the number of leading points.Besides，registration accuracy is improved.

Key words: attribute vector, MODIFIED HAMMER algorithm, feature_map, ultrasound image registration