Inhomogeneous illuminated images registration based on wavelet decomposition
Bing Luo, Junying Gan · 2009
In PCB SMT assembly products automated machine vision inspection, images registration is necessary for getting an entire PCB image. Because of inhomogeneous illumination in engineering images capturing, conventional mosaic method can not deal with it. High frequency coefficients of wavelet decomposition reflect the contour of the image and that of illumination variety is relative small. So images registration based on wavelet decomposition can be robust for machine vision inspection. Using projection of wavelet decomposition coefficients to instead of mutual correlation can simplify the calculation from 2D to 1D. Experimental results show this approach is effective, robust and quick.