A new method based on gold image subtraction (GIS) combined with the Fourier transform displacement theorem (FTDT) curve matching algorithm was proposed to detect the defects of printed fabrics caused by color and pattern diversity. First, a Gaussian filter was applied to printed fabrics to eliminate the influence of noise on the defect detection results. Then, GIS was employed for image matching and defect segmentation. Finally, a curve matching algorithm based on FTDT was adapted to calculate the relative offset of the cross-printing defects. Experiment results indicate that the algorithm can accurately segment the defect and provide detailed information about the defect. The algorithm can also accurately obtain the relative offset of cross-printing defects, which has a certain guiding significance for industrial production.