图像配准是机器视觉和图像处理的重要研究内容, 针对印刷缺陷检测中图像配准效率低、误差大等问题, 提出了区别于普通连续色调图像的新方法。采用数学形态学和Hough变换的方法检测图像配准的旋转角和缩放因子, 并基于相位相关原理对傅里叶变换的范围进行改进, 求取待检图像和标准图像之间的相对偏移量, 最终实现文本印刷图像的高效配准。实验结果表明, 平均配准误差控制在1个像素范围以内, 而且整个配准过程在1s内完成, 对文本图像具有较好地配准效果, 能为文本印刷品的质量检测奠定良好的基础。

Image registration is important research content of machine vision and image processing. A new idea and method is proposed aiming at the problem of text image registration in printing detection. The rotation angle and scale are detected by the method of mathematical morphology and Hough transform. The relative offset between the image and the standard image is obtained by using the improved local phase correlation method, and finally the text image registration is realized. The experimental results show that the method has high accuracy, strong robustness, fast speed and better registration effect for text image, which can lay a good foundation for the quality inspection of the text printed matter.