Abstract: This paper analyzed some existing image watermarking algorithms using
singular value decomposition(SVD) technique, and pointed out they have the problem of high error detection rate of original image and randomly chosen images. With the analysis of SVD's properties a new algorithm was proposed. Experiments showed the new proposed algorithm can distinguish the watermarked image from unwatemarked original image.It can also
significantly decrease error detection rate of randomly chosen images.
This algorithm has good robustness.

Key words: SVD, error detection rate, counterfeit watermarking, most important column, digital watermarking, SVD, error detection rate, counterfeit watermarking, most important column