一种具有旋转不变性的相关系数影像校正方法

Correlation coefficient image correction methods with rotation invariance

相关系数校正法对影像旋转敏感度高,不具有旋转不变性。针对此问题,提出一种基于霍夫变换和1D傅里叶变换的灰度相关系数影像校正方法。首先将影像进行霍夫变换和傅里叶变换,将旋转问题转变成平移问题,利用相关系数法求出其平移参数,即原始待处理影像的旋转参数;将待处理影像去旋转后,再运用相关系数法进行平移校正得到最终的校正影像。实验结果表明此方法能准确有效快速地处理具有旋转和平移的影像校正问题。

The correlation coefficient correction method has high sensitivity to the image rotation,but it does not have rotation invariance. According to the theory,this paper proposed a new one based on Hough transform and 1D Fourier-transform. Firstly,it used these two transformation to turn the problem of rotation into translation,then obtained the rotating angle of original image by the traditional method. After de-rotation,it used the traditional method in translational correction so that the final corrected image could be got. Experiments show that the proposed method can detect the rotation and translation between images efficiently,exactly and quickly.