In order to solve the problem of the sensitivity of the highest singular value bit under geometric attack and the poor anti-geometric attack performance of zero watermark, an image zero watermark technology based on ameliorated singular value and sub-block mapping is proposed. Firstly, the carrier image was Arnold scrambled to eliminate the correlation between pixels, and then Curvelet transformation, partitioning and singular value decomposition were performed to obtain the stable characteristics of the carrier image. At the same time, a seed block mapping mechanism is proposed to divide the original copyright image into different sub-blocks, and by suming the sub-blocks, different characters are used to represent the water imprint block. Finally, the feature of the carrier image and the water imprint block are logically calculated to generate zero watermark by bit. Experimental results show that the proposed zero-watermarking algorithm has strong robustness under geometric attack, non-geometric attack and combined attack, and the generated zero-watermark information is more secure.