融合SIFT的B样条红外图像非刚性配准

手持式广角镜头红外热像仪所拍摄的不同时刻红外图像具有刚性形变和非刚性形变,传统图像配准算法很难同时矫正刚性形变与非刚性形变,针对该问题,提出一种融合SIFT的B样条配准算法。首先在待配准图像中建立控制网格,其次运用SIFT算法寻找待配准与基准图像间的匹配点对,剔除错误匹配点对并计算出待配准图像与基准图像间的刚性变换参数,接着对控制点进行刚性变换,最后以局部强度和为测度函数,运用B样条非刚性配准算法对广角镜头引起图像的非线性进行矫正。对比实验结果表明,本文算法具有很高配准精度,能够满足实际工程精度要求。

B-spline non-rigid registration algorithm for infrared image based on SIFT

Infrared images obtained by handheld infrared camera with wide-angle lens have rigid and non-rigid deformations. Tradition image registration algorithm is difficult to correct the rigid and non-rigid deformations. To solve this problem,B-spline non-rigid registration algorithm based on SIFT is proposed. At first,the control mesh is created in the input image. Then,matching points between input image and template image are found by SIFT algorithm. The rigid transformation parameters are calculated after ignoring the incorrect matching points. Next,the control points are transformed by rigid transformation parameters. Finally,the sum of pattern intensity is used as measurement,and the nonlinear transformation of the image that is caused by wide-angle lens is corrected by B-spline non-rigid registration algorithm. The results of comparison show that the new method has better registration accuracy and it also can meet the requirements of practical engineering accuracy.