ABSTRACT
Lossless data hiding technology aims to embed secret data into a digital multimedia signal in a reversible fashion, also called reversible watermarking [1]. Different from traditional robust watermarking scheme, for reversible watermarking technology, at the decoder, both the original cover image and embedded secret data need to be completely recovered and extracted from the watermarked image. Recently, many reversible watermarking algorithms have been presented in the literature, e.g. the schemes based difference expansion [2], [3], [5], histogram shifting [1], [4], [6]-[8], etc. In order to increase the embedding capacity, Al-qershi et al. proposed a novel scheme based on two-dimensional difference expansion technique instead of conventional difference expansion [5]. This algorithm uses Haar discrete wavelet transform to transform each block into frequency domain. If the block belongs to expandable block, each coefficient of HH1 sub-band can embed two bits secret data, and each coefficient of LH1 and HL1 sub-band can embed one bit. So, it can provide 16 bits per block. However, although the embedding capacity is improved, the PSNR of this scheme is quite low, and it does not do comparison with multi-level scheme.
