A robust zero-watermarking algorithm resisting JPEG compression and geometric attacks is proposed, since digital images always face with JPEG compression and geometric attacks through transmission. An original image is split into non-overlapping blocks, and each block is conducted with singular value decomposition.Then, each block′s singular value matrix is transformed with Harr wavelet transformation. The zero-watermark sequence is produced by judging the numerical relation between the mean of diagonal elements of singular value matrix′s low frequency band from two adjacent blocks. The mathematical theoretical analysis shows that the zero-watermark sequence is essentially produced by judging the numerical relation between the mean of singular value matrix′s singular values from two adjacent blocks. The proposed algorithm has perfect visibility due to no alteration made to the original image. The experimental results show that the proposed algorithm has strong robustness to resist JPEG compression and several geometric attacks such as rotation, scaling, random row & column removal, row & column shifting, print-and-scan.