A HIGH-CAPACITY REVERSIBLE DATA HIDING SCHEME BASED ON PAIRWISE PEE AND EMD

Abstract

This paper proposes a high embedding capacity reversible data hiding algorithm using pairwise prediction error expansion and exploiting modification direction. In this algorithm, the original image is first decomposed into 4 × 4 blocks, and each block is divided into inner and outer sub-blocks. Next, the standard deviation (δ) of the pixels in the outer sub-block is used to evaluate the statistical properties of the inner sub-block. If the value δ of the outer sub-block is less than a given threshold, the inner sub-block is considered embeddable, otherwise, it is labeled non-embeddable. A pairwise prediction error expansion mapping is adopted for embeddable inner sub-blocks. Additionally, to improve the embedding capacity of the proposed scheme, an exploiting modification direction table is developed to modify the prediction error histogram. Through careful selection of error pairs for mapping using the exploiting modification direction table, the quality of the stego image can be preserved while the embedding capacity of the proposed method can be significantly improved. Experimental results illustrate that our method achieves better performance than previous works, especially in terms of image quality.