Adaptive Self-Embedding Scheme With Controlled Reconstruction Performance

In this paper, we address the problem of adaptive self-embedding, where the reconstruction quality is controlled individually for different fragments of a digital image. We focus on the impact of incorporating content adaptivity features on the restoration success conditions and the achievable reconstruction performance. We analyze the problem theoretically and validate the obtained results experimentally with a fully functional selfembedding scheme. Our analysis shows that introduction of multiple reconstruction profiles, even with significantly lower restoration fidelity, does not need to improve the achievable tampering rate bounds. The obtained fine-grained control over the reconstruction process is exploited to provide guarantees on certain performance aspects. Based on the derived theoretical model, we propose a procedure for optimization of the overall reconstruction quality given constraints on the desired target tampering rate and the required quality level for selected image fragments. Such guarantees are of principal importance in a number of applications.