Automatic density measurement of yarn dyed woven fabrics was discussed. The dual-side images were captured using the lab-used dual-side image acquisition system. Radon transform and affine transform were utilized to align and match the dual-side images. The dual-side images were fused based on two kinds of image fusion methods, and the frequency points corresponding to the periodical structure could be extracted with the aid of Fast Fourier Transform (FFT) technique. Meanwhile, the reconstructed images could be obtained based on inverse Fast Fourier Transform (IFFT) technique, and the yarn density could be measured. The power spectrums of one-side image and dual-side fused image were analyzed. It is found that the texture of one-side image is not continuously connected and the frequency spectrum is scattered. However, the spectrum energy of fuses image focused on the horizontal or vertical direction so that the corresponding frequency points could be easily extracted. The test result shows that the maximum deviation between our method and the density meter based method is 1.96%, and the average deviation is 0.86%. It is considered that the method could improve the accuracy of density measurement of yarn dyed woven fabrics.