A New Pixel to Block Mapping for Efficient and High Fidelity Data Hiding

Reversible Data Hiding (RDH) with Pixel-to-Block (PTB) mapping has recently been studied for content authentication and annotation applications of digital images keeping the original pixels intact in an embedded image. This paper presents a new PTB mapping technique and its uses in developing an RDH scheme. Existing PTB techniques generate an image block of size 2×2 from an original pixel using a predefined predictor. For embedding 4-bits of secret data, pixels of each block are then shuffled to form a unique combination of pixels for each value of the 4-bit binary data. In contrast, we propose to compute a 2×2 block copying an original pixel to all four pixels. The first pixel is kept unchanged and the others are modified to construct a unique block mapped to the values of 4-bit data. With the proposed minimum possible changes in the three pixels of a block would offer a better embedded image quality. The effective embedding capacity would also be increased as such the proposed PTB technique can tackle the overflow and underflow problem without any location map. Early experimental results of the proposed RDH scheme with the new PTB mapping recorded a significantly better rate-distortion performance, while compared to the state-of-the-art RDH schemes.