A Quantization Watermarking Technique Robust to Linear and Non-linear Valumetric Distortions Using a Fractal Set of Floating Quantizers

Abstract

This paper presents an extension of the classical Quantization Index Modulation (QIM) data-hiding scheme in the context of valumetric distortions. This scheme uses a fractal quantization structure during the detection but also a content dependent quantization grid to achieve both global constant robustness and the ability to recover the watermark after non-linear valumetric distortions. Previous works are first presented. Then the construction of a floating quantizer that addresses the problem of non-linear transformations is introduced. The embedding and detection schemes for digital image watermarking are afterward introduced, the main characteristic of this scheme is the fact that the detection scheme can use a hierarchical set of quantizers to deal with non-linear valumetric transforms while preserving an average constant quantization step. Finally the performance of this scheme and the comparison with other robust quantization schemes considering valumetric transforms and noise addition are presented.