Efficient reversible data hiding in encrypted binary image with Huffman encoding and weight prediction

Abstract

This paper proposes an efficient reversible data hiding (RDH) scheme in encrypted binary image by combining Huffman encoding and weighted prediction. Since the pixel redundant space in binary image is very small, existing RDH schemes for encrypted binary image is very difficult to get a balance between large embedding capacity and high visual quality of recovered binary image. To provide an efficient solution, we first divide the original binary image into three different types of non-overlapping blocks, black blocks, white blocks and mixed blocks. Subsequently, the Huffman encoding mechanism is introduced to achieve a large embedding capacity in all blocks. Due to the correct decodability of Huffman coding, both black blocks and white blocks can be recovered lossless after extracting secret data. Furthermore, a weight prediction mechanism is designed to make an accurate prediction for the mixed blocks, which can be sequentially recovered by involving a large area of pixel correlation. Our method can efficiently achieve a large embedding capacity for additional data, while keeping a high visual quality for recovered image. Extensive experiments demonstrate that our method outperforms existing encrypted binary RDH schemes with higher visual quality and larger embedding capacity.