Research On Stereoscopic Image Fragile And Semifragile Watermarking Methods And Their Applications

Stereoscopic image/video system provides viewers the immersive sense. Nevertheless, thesecurity is one of the urgent issues to be tackled for promoting the entertainment industry.Stereoscopic image fragile and semifragile watermarking methods which are intended for contentauthentication and integrity verification, as well as the quality assessment, are investigated deeplyinthisthesis,andtheworkconsistsofthreeparts as brieflysummarized below.To improve the quality of watermarked image, ability of self-recovery and accuracy of tamperlocalization, four fragile watermarking methods with self-recovery capability are proposed toaddress the problem of few efforts devoted to stereoscopic authentication.1) In the first method,restoration-bits are generated from disparity and the roughness of stereoscopic images. Two copiesof the part restoration-bits are embedded into the image, the disparity is also used to recover thematched block contents, and consequently the high restoration quality is obtained.2) Two fragilewatermarking methods are proposed based on joint just noticeable distortion. The watermark of thetwo methods is generated from the frequency and spatial domain, respectively. Experimentalresults show that the two proposed methods provide high restoration quality and precise accuracyof tamper localization.3) Alterable-length fragile watermarking method for stereoscopic imageauthentication is proposed with stereo matching technique. In this method, the restoration-bits aregenerated from the frequency or spatial domain according to the roughness of image. Disparity isused to recover the matched contents so that the watermark capacity is decreased observably. Then,the watermark is divided into three parts and embedded into three mapping blocks, and thewatermarked and recovered results with high quality are obtained. A detection technique with thepart of alterable-length watermark is used to localize tamper with the accuracy of tamper detectionofhigher than99.5%.To resolve the drawback that the semifragile methods do not achieve high tampered detectionand the human visual perception characteristics are not considered in the process of watermarkembedding, two stereoscopic semifragile methods are proposed.1) The first semifragilewatermarking method is based on singular values and quantization in the wavelet domain. Thefrangibility of watermark is adjusted by the quantization step. The watermark constructed by theproposed method is robust to mild attacks and fragile to malicious attacks, and achieves preciseaccuracy of tamper localization.2) Binocular just-noticeable-difference based stereoscopic image watermarking method is proposed with self-recovery. Firstly, authentication-bits are generated andembedded by using singular values with its strong stability and binocular just-noticeable-differencevalue. The quality of watermarked image and restoration is improved by taking full advantage ofthe matching characteristics between the left and right images, and the accuracy of tampereddetection ishigherthan98%undermaliciousattacks.To improve the accuracy of quality evaluation method based on watermarking and thetransparency of embedded watermark, an image quality evaluation method based on joint justnoticeable distoration and quantization technology for JPEG compression is proposed. Thewatermark is embedded into the complex regions, and the quantization step is adjusted usingfeedforward and feedback control combined with joint just noticeable distoration and full objectivequality assessment model. Therefore, the evaluated quality generated by the degradation of theextracted watermark is consistentwith thehuman visual perception. Experimentalresults showthat,by applying the proposed model to stereoscopic test database, Person’s correlation coefficientsindex reachesto0.9108,Rankedcorrelationcoefficientsindex reachesto0.9261.