Less is more: learning predictability in reversible steganography

Abstract

Artificial neural networks have advanced the frontiers of reversible steganography. The core strength of neural networks is the ability to render accurate predictions for a bewildering variety of data. Residual modulation is recognised as the most advanced reversible steganographic algorithm for digital images. The pivot of this algorithm is predictive analytics in which pixel intensities are predicted given some pixel-wise contextual information. This task can be perceived as a low-level vision problem and hence neural networks for addressing a similar class of problems can be deployed. On top of the prior art, this paper analyses predictability and endows the predictive analytics module with the option to abstain when encountering a high level of uncertainty. Predictability analysis can be formulated as a pixel-level binary classification problem and tackled by both supervised and unsupervised learning. In contrast to conventional statistical analysers, a learning machine can automatically acquire some statistical principles regarding image patterns and simultaneously adapt to a specific predictor. Selective prediction induces fewer and yet predictable pixels designated to carry the payload, which in turn leads to better capacity–imperceptibility performance, reflecting the ‘less is more’ philosophy. Experimental results show that steganographic performance can be remarkably improved by adaptively filtering out the unpredictable pixels with the learning-based predictability analysers.