At present, the digital image watermarking algorithm has not embedded the synchronization signal in the image, resulting in the poor performance of the embedded image in terms of, for example, imperceptibility, anti-attack ability, and robustness. Therefore, a digital image watermarking algorithm based on scrambling and singular value decomposition is proposed. The digital watermark is preprocessed by dimensionality reduction and encryption; the digital image is processed in sections and embedded in the synchronization signal. The digital watermark image is embedded in the digital image watermark by using the low-frequency energy ratio technology of sound channel. The watermark image extraction step is designed to extract the watermark image. The transmittance of the degraded image is calculated, the transmittance is refined by relying on the soft matting algorithm and guiding filter, the image contrast in the image frequency domain is enhanced, and the results are mapped to the appropriate visual range to optimize the visual brightness of the image. Finally, according to the uniqueness of singular value matrix, the distributed characteristics of image matrix data are described, the visual effect is enhanced, and the research of digital image watermarking algorithm is completed. The experimental results show that the algorithm can extract the watermark information completely, the image contrast and singular value have been significantly improved, and the algorithm has better anti-attack performance.