Research On High Capacity Reversible Watermarking Algorithm For2-D Code Applications

Military and medical images, which are images with high demands of accuracy,often need to be embedded authentication information. Some of the information will bedestroyed permanently if traditional digital watermarking algorithm is used. But a smallmistake will cause a big problem in these areas, so traditional digital watermarkingalgorithm cannot be used. Reversible watermarking is a hot research area in recent years,which can not only extract the watermark information but also recover the originalimage precisely. These features can satisfy the requirements of such applications.QR code is a popular2-D code in recent years. It has several great features, such asfast recognition, high fault-correctness rate, and large capacity. QR code application onimages has become more and more frequent. However, if the QR code is used on animage, a part of the image will be covered so that the full image can not be recovered.In order to precisely recover an image with QR code just using the image itself,two reversible watermarking algorithms with high capacity are proposed, which hasgood improvements on the existing watermarking algorithm. Base on this, the thesisalso proposes an algorithm that embeds the covered image part by QR codes into theoriginal image as a part of watermark information so that the full image can berecovered accurately. In this thesis, the main work is as follows:①A novel reversible watermarking algorithm based on difference expansion isproposed. The algorithm has large capacity and good multi-rounds embedding capacity.Based on the deepstudy of traditional watermarking algorithms, only a pair of pixels isused to embed two watermark bits. The embedding capacity can reach1bpp (bits perpixel); while other current best known algorithms can only reach0.9bpp. It also has agood performance on embedding capacity improvements for multi-rounds embedding.②A new histogram shifting reversible watermarking algorithm based on discreteHaar wavelet transform coefficients is proposed. Different from the traditional grayscalehistogram shifting reversible watermarking algorithm, the new proposed algorithm usesthe discrete Haar wavelet transform coefficients of the image pixels to generatehistogram. The capacity of the algorithm can be improved because these coefficientswill be more focused on ’0’ point after transforming. Besides, the algorithm overcomes adrawback of the traditional histogram shifting algorithms that they could not deal withthe image with the balance gray histogram. ③The two above reversible watermarking algorithms are applied to the imagesthat contain QR codes. A method is proposed to embed the image block part covered byQR code into the image itself. At the extraction side, the image can be recoveredprecisely without using other auxiliary information. Besides, the watermark can also beembedded to fulfill other security requirements.