基于纹理公因子的异源图像配准方法

摘要

异源图像的成像机理不同，导致同一场景的异源图像之间存在着较大的形变或者光照等差异，异源图像配准很难直接运用尺度不变特征变换算法（SIFT）和快速鲁棒特征算法（SURF）等方法。为此，针对异源图像提出一种基于纹理公因子的配准方法。根据傅里叶变换将异源图像变换到频率域并用 Gabor模版进行滤波处理，在空间域中利用 Sobel算子对异源图像进行纹理公因子提取，采用自定义规则选择匹配点，通过随机抽样一致性算法对匹配点对进行提纯，根据提纯后的匹配点对求解单应变换参数，经过坐标变换及插值分析实现异源图像配准。实验结果表明，与 SIFT和 SURF等算法相比，该方法的匹配准确率较高，鲁棒性和异源图像配准效果较好。%The difference of imaging mechanism in different source image leads to large deformation or light and other differences between two different source images in the same scene,so it is difficult to directly use Scale Invariant Feature Transform(SIFT)or Speeded Up Robust Features(SURF)for different source images registration.For this reason,a registration method based on texture common factor is proposed for different source image.According to Fourier transform,the different source image can be transformed to frequency domain,and Gabor template is used for filtering. The method uses Sobel operator to extract the texture common factor from different source image in spatial domain. Matching point is selected by a custom rule,and the matching points can be purred by Random Sample Consensus (RANSC).According to the purified matching points,the method can obtain the homography transformation parameters. Finally,after coordinate transformation and interpolation analysis,it achieves registration between different source images. Experimental results show that the method proposed in this paper has high matching accuracy and robustness compared with SIFT and SURF.