Reversible medical image watermarking based on wavelet histogram shifting

This paper introduces a reversible and scalable blind watermarking method for medical images based on histogram shifting in wavelet domain. The histogram shifting-based watermarking especially in spatial domain suffers from the overhead of positions information that has to be embedded. This not only has negative impact on the capacity of the embedded data, but also reduces the quality of the watermarked image. To overcome this problem, a new histogram shifting approach in wavelet domain is introduced which adaptively manages the parts of the required histogram to be shifted. For inserting watermark data, two thresholds, T-1 and T-2, are determined in the high-frequency sub-bands of the histogram based on the size of watermark data. Two zero-points, Z(1) and Z(2), are created in the histogram shifting. In the histogram shifting procedure, only small parts of the histogram have been shifted and therefore, introduced distortion has been considerably decreased and the quality of the watermarked image has been increased. Applications of the method were examined for the multiple medical image watermarking. Two sets of data, one as copyright watermark and the other as caption watermark which contains patients' information, are first encrypted and then inserted in the high-frequency sub-bands in the integer wavelet domain. Inserting the data in different subbands besides encryption of it by a private key provides high security for the embedded data. Experimental results obtained for several watermarked medical images indicate imperceptibility of the approach even for high payloads. The proposed approach overcomes other related works introduced in the recent literatures.