A near-lossless data hiding method with an improved quantizer

This paper proposes a method for near-lossless data hiding (DH). A DH method distorts an image to embed data in the image, and the method extracts the hidden data from the distorted image. The proposed method also recovers an image from the distorted image where a pixel in the recovered image differs from that in the original by not more than a pre-specified value, i.e., near-lossless. First in the proposed method, an image is quantized by an improved nonuniform quantizer to prepare room for DH. Then, the histogram-based lossless DH technique conceals data to the quantized image. The proposed method is superior to the conventional near-lossless DH method in terms of the amount of the hidden message, whereas both methods use the same pre-specified value. Moreover, the proposed method also improves the quality of the recovered image in some conditions.