The work aims to propose a set of methods to eliminate the pseudo-defects by improving the registration accuracy and spatial filtering to reduce false inspection during automatic optical inspection (AOI) of printing product by difference image method. Firstly, the registration accuracy of contour-based template matching was improved by ant colony optimization for continuous domains for global optimal solution to reduce the pseudo-defects. Secondly, the image was cut into two parts: contour zone and non-contour zone according to the distribution characteristics of the artifacts caused by affine transformation during automatic registration. After weakened by corresponding spatial filtering, the pseudo-defects were eliminated with threshold segmentation. Under experimental environment, the contour-based template matching improved by ant colony optimization for continuous domains could be accurate to the sub-image level, and the registration rate was 92%. The false inspection rate was 0 and the missed inspection rate was 3% after the pseudo-defects were eliminated with spatial filtering. The average inspection time was 1.05 s and the maximum was less than 1.5 s. The proposed method improves the effect of image difference in AOI and can be realized fast, effectively and easily. It reduces the false inspection rate of defects without causing missed inspection and can be used in actual industrial production while meeting the requirements of online inspection.