基于小波变换与相位相关的PCB图像拼接算法*

在印刷电路板自动光学视觉检测(automatic optic inspection,AOI)中,需要通过图像拼接以获得整个PCB的图像。传统的图像拼接算法处理光照干扰效果不好，并且算法复杂。针对AOI的工程实际，通过对图像进行二维小波变换去掉反映光照渐变的低能量系数，获得反映图像水平和垂直方向变化轮廓的高频系数,同时有效地降低计算复杂度；结合相位相关法实现快速精确的重叠位置配准，采用加权平均法实现图像的融合。实验结果证明，该算法能够对多种规格、不同颜色的PCB图像实现快速、准确的拼接。

PCB image mosaics algorithm based on wavelet transformation and phase correlation

In order to stitch images captured by camera into an entire one, this paper developed an image mosaic method. Traditional image mosaics algorithms did not work well on AOI because of optical noise and complexity. It used two-dimensio-nal wavelet transformation to get the high frequency coefficients that reflected the changes of image in horizontal and vertical direction, at the same time it removed the low coefficients that reflected the gradual changes of illumination and reduced the computational complexity. Then the algorithm used phase correlation technology to determine the location of the overlap, and achieved image fusion by weighted average method. The proposed algorithm was tested by experiments and worked well.