中文摘要：

为提高图像配准方法中描述子方向的计算精度和效率,设计基于亮度质心不变矩的轴向密集插值法和面积积分法。配准前,将复杂计算融入系数矩阵中,实现配准实验系统。实验结果表明,两种方法的计算速度均比SURF快4倍;面积积分法的角度计算精度在全方位上比SURF算法和轴向密集插值法更加均衡,对实验图像配准的重复率也达到SURF水平;轴向密集插值法在45°方向附近有着远高于SURF算法的表现,但在其它角度上表现不稳定,提示了通过优化系数矩阵提高方向矢量精度的可能性。

英文摘要：

To improve the accuracy and efficiency of the descriptor orientation calculation in image registration method, axial concentrated interpolation method and area integration method based on the brightness centroid invariant moment were designed. The methods completed the complicated calculation as the coefficient matrix before registration. The registration test system was designed. The experimental results show that the calculation speed of the two methods are 4 times higher than that of SURF （speeded up robust features）. The angle calculation precision of area integral method is more balanced than the SURF algorithm and the axial intensity interpolation method in all directions, and the repeatability of it is as good as that of SURF through the test image registration. The axial intensity interpolation method has much better performance than the SURF algorithm on 45° orientation, which suggests the possibility to improve the accuracy of the orientation by optimizing the coefficient matrix, hut it is not steady in other directions.