Reversible watermarking based on interpolation error histogram shifting

Abstract

A reversible data hiding technique with high capacity of data embedding is presented in this paper. Using both difference expansion and histogram shifting methods at the same time we have improved Luo et al.'s method to increase the amount of data that can be embedded into the host image. First the image is examined for possible overflow/underflow problem that might occur during embedding; then if needed, histogram is shifted to prevent this overflow/underflow. At next stage, interpolation error is calculated and histogram of the interpolation error is shifted to embed watermark data. Because overflow/underflow is prevented, no boundary and/or overflow map is needed and instead a much smaller location map is used so that we can fully recover the image in extraction phase. Also, overhead data is embedded into the host image along with the actual watermark; therefore, no extra data is needed to be transmitted to receiving side as needed for some reversible watermarking schemes. Experimental results show that this method can achieve higher data embedding capacity compared to other reversible watermarking schemes at same PSNR.