Texture driven adaptive multi-level block selection based reversible data hiding in encrypted image

For reversible data hiding in encrypted image (RDHEI), it is an important problem to resolve the contradiction between embedding capacity and image visual quality. However, existing schemes do not maintain a good balance between them. Accordingly, we propose a multi-level block selection embedding mechanism to solve this problem. Our scheme firstly divides original image into muliple non-overlapping blocks with identical size. The most significant bit (MSB) planes of all pixels in the block are predicted from their neighboring pixels to generate a corresponding prediction error map for each block. Subsequently, the pixels are divided into eight bit planes, and the adjacent bit planes are implemented XOR operation to vacate room. Furthermore, by simulating data embedding in each block, PSNR values of each image block are sequentially generated. After encrypting the image using stream ciphers, the optimal blocks with maximum PSNR value can be selected to embed data according to the secret data capacity. Since our proposed scheme usually selects the image block with maximum PSNR value to embed secret data, it can naturally improve the image visual quality. Extensive experiments demonstrate that our scheme is effective and efficient, and outperforms existing state-of-the art methods with better visual quality and higher security performance under the given embedding capacity.