二维点模式图像的仿射变换配准

提出了一种图像特征点配准算法．该算法为二维点模式图像仿射配准建立了新的求解模型，此模型对参考图像的特征点集和浮动图像的特征点集分别进行Whitening变换，将点集间的一般仿射变换问题转换为刚性变换问题；在对刚性变换求解时，采用平滑性好、局部极值较少的新的目标函数；并引入了形变程度分量，使该算法更能符合实际应用．结合文中提出的新的仿射不变量，目标函数只需在平分法的基础上加入随机因素便能快速求解．实验结果证明，该算法在处理特征点仿射配准问题上具有速度快、精度高的特点．

Affine Registration of 2D Point Pattern Image

This paper presents a novel algorithm for the problem of affine point pattern matching. After using a Whitening transformation, the two point patterns related through an affine transform are respectively mapped to the point patterns related through a rotation transformation. The novelty of the proposed algorithm is by introducing a new distance metric and a distortion degree of the affine transformation into the objective function to make it smooth and have relatively less local minimum. Combined with the newly proposed robust affine invariant, the objective function can be solved rapidly by random bisearching algorithm. The experimental results show that the proposed algorithm is capable of affine point matching with higher speed and precision.