Dual-prediction and adaptive complexity for reversible watermarking of color images

To expand the application scope of digital watermarking for color images widely used in social communication, this paper proposes a reversible watermarking scheme for color images. This scheme integrates bidirectional prediction correction embedding, dual dimensional complexity evaluation, and multi-threshold joint control to achieve adaptive optimization prediction of embedding parameters. First, by defining reference and associated channels, bidirectional correction is achieved through forward utilization of the reference image to predict and embed data into associated images, while simultaneously utilizing associated images backward to optimize and correct points with large disturbances during the embedding of the reference image. Second, a dual-dimensional complexity evaluation model is constructed by fusing local variance and visual saliency features, which accurately characterizes both macroscopic edge structures and microscopic texture details of images to precisely localize optimal embedding regions. Finally, multi-threshold joint constraints facilitate adaptive selection of optimal embedding sizes, mitigating distortion induced by capacity expansion and invalid pixel shifts. Experimental results demonstrate that the proposed scheme outperforms state-of-the-art methods, adapting better to the content characteristics of color images and achieving a superior balance between image quality and embedding performance.