Research On Digital Watermarking For Stereoscopic Image/Video

Stereoscopic percetion and real scenes feeling are emerged from the stereoscopic videosystem. The stereoscopic video system has broad application prospects. With the developmentof technology, these high-end applications would be copied, tampered and attacked in otherforms, which resulting the product owners unprotected. How to research rational and effectivestereoscopic video of the watermarking using the unique characteristics of stereoscopicimages is a worth exploring and in-depth research new topic. In this paper, the watermarkingtechnology used in stereoscopic video image quality evaluation and copyright protection isresearched based on the watermarking technology of two-dimensional images by analysingthe unique features of the stereoscopic video images.First, a number of watermarking algorithms for stereoscopic image are proposed byanalyzing some of the existing two-dimensional video images digital watermarkingalgorithms combined with the unique characteristics of stereoscopic video images. Thesealgorithms are suitable to the processing of basic stereoscopic video image such as JPEGcompression, filtering, etc. And these algorithms reflect the robustness from different anglesto achieve copyright protection of stereoscopic video image from infringement. Among these,the watermark information is embed using the relationship between AC coefficient and theenergy in the block of stereoscopic image in the proposed blind watermarking algorithmbased on the modulation of relationship, the robustness of the algorithm is determined by theembedded location of choice and the established relationship between the left and right image.The proposed watermarking algorithm utilizing the joint relation and the quantification forstereoscopic image uses the stability relationship between the blocks of stereoscopic image toselect the embedding method. It improves the robustness and the transparency of watermark.The same nature block with the same change trend in left image and right image is combinedin the proposed blind watermarking algorithm based on the stereoscopic image feature, andthen the way of selecting embedded method is also used to improve the contradictionsbetween algorithm robustness and watermark transparency in this algorithm.Second, summing and analyzing the feature of the above algorithms, the quantitativestudy for stereoscopic video image robust zero-watermark technology based on combinedwith the advantages of two-dimensional video images zero-watermarking technology is given.According to the disparity relations of stereoscopic image, a stereo image zero-watermarkingmethod based on block matching is proposed. The left image block is matched from the rightimage using disparity, and then the zero-watermark is composed by the relationship between the matching errors of the best match and the average error of overall. Experimental resultsshow that the method with a large enough threshold can effectively achieve the purpose ofcopyright protection of stereoscopic images, and it is robust to the basic symmetric andasymmetric distortion process (such as JPEG compression, filtering and cropping) of thestereoscopic images. Based on the strong stability of SVD decomposition, thezero-watermarking algorithm based on disparity matching with SVD decomposition isproposed. Compare with the fist method, it has more robustness. There is a strong correlationin stereoscopic image. According to this, a zero-watermarking method of stereoscopic imagebased on the correlation in view-point blocks is proposed. The correlation is defined on theeigenvectors of Single Value Decomposition (SVD) transformed blocks in left and rightimages. The zero-watermark is generated by comparison between the correlation of pairblocks and the average correlation of all blocks pairs. Experiments show that the proposedmethod has a good robustness against the basic image processing such as JPEG compression,filtering and cropping. In order to improve the robustness of the proposed method, blocks ofboth images will be classified into two texture types including flat block and non-flat block.According to the texture information of blocks, the left and the right block in the sameposition will be paired and assigned to the strong correlation pair or the weak correlation pairgroup. A zero-watermark method based on hierarchical correlations between the view-pointof blocks for stereoscopic images is proposed.Then, by analyzing the human eye’s sensitivity to different regions of different texture,the image texture is classified as: flat blocks, edge blocks and texture blocks, and theclassifications are according to the relationship of energy in image block after DCTtransformation. Combined with Watson perceptual model, a digital watermarking algorithmfor image quality assessment is proposed. The different intensity image watermark isembedded in different blocks according with the Watson model and human visualcharacteristics in this method. The experimental results show that the quality assessment ofJPEG compression images and subjective quality of these images are in good agreement. Inthe realization of this new application, using the advantages of zero-watermark and combingwith the different texture masking in stereoscopic images, a digital watermarking algorithmfor stereoscopic image quality assessment is proposed. The algorithm not only reflects thedifferent texture structure in the human eye’s perception with the different subjectiveevaluation, but also reflects the different depth of stereoscopic in human eye’s perception withthe different degree of subjective evaluation.Finally, the different structures of the stereoscopic video coding performance aresummarized and analyzed, a perceptive watermarking algorithm for stereoscopic video isproposed. The watermark information is embed by adjusting the QP of for the right viewvideo, while there are two ways to adjust the QP in this perceptive video watermarkingmethod, one is to adjust the QP of the whole frames in the right view video, and the other is toadjust the QP of the right view video image block. The right view video image block is classified, and the QP is different in different type image blocks. Experimental results showthe different watermarking method will be used in the different encoding conditions.