Research On Authentication Techniques For Depth-image-based Rendering 3D Image

With the higher demands for the three-dimensional sense and realistic of image and video,free viewpoint technology emerged.As a key technology of free viewpoint,virtual viewpoint rendering technology can generate the object information of new viewpoint,providing the users with the depth information of the object,which can lead user to a stereo world.Depth-image-based Rendering(DIBR)is considered as a kind of typical virtual view rendering technology.DIBR has a lot of advantages,including singleness of data source,low requirement for transmission bandwidth,rendering with high speed and high quality of image.With the characteristics above,it has received abroad attention and research,developing at very fast speed.As a matter of fact,DIBR 3D images(videos)could be distributed and copied illegally,the authenticity and integrity of contents should be verified and protected.This thesis discusses the multimedia security issues of DIBR 3D images.With the study of the traditional digital image authentication technologies,we propose robust watermarking technology and robust hashing technology for DIBR 3D image,aiming to implement the digital image authentication and achieve the purpose of copyright protection and content protection.In this thesis,we have made the following contributions:(1)Taking account of the special requirement of robustness for DIBR 3D image watermarking,we propose two kinds of DIBR 3D images watermarking algorithms in transform domain using spread spectrum and quantification of wavelet coefficients,respectively.As experimental results shown,the proposed watermarking algorithms are robust to DIBR process(i.e.the watermark embedded in the center image can be blindly extracted from the virtual images).The proposed methods are robust to common signal distortions.In addition,the proposed methods are robust to baseline distance adjusting and depth image blurred.(2)To address the problem that existing DIBR 3D image watermarking algorithms are sensitive to geometric attacks,we propose a robust blind watermarking method based on histogram shape to protect DIBR 3D images.As the experimental results shown,the proposed method is robust to baseline distance adjusting and depth image blurred.Benefit From the geometrical invariance of histogram shape,the proposed method is much more robust to geometric attacks.(3)Taking account of the special requirement of robustness for DIBR 3D image hashing,we propose a robust DIBR 3D image hashing algorithm based on translation invariance of image.Using the characteristics of image gray histogram geometric invariance,we design a robust hash algorithm resisting to geometric attacks.Experiments show that the proposed hashing is not only robust to DIBR process,but also robust to the signal distortion attacks such as compression,adding noise and filtering.Meanwhile,we extract the DT-CWT coefficients under different gray-scale of the image to makes the hashing algorithm sensitive to localized tampering.(4)In view of the special application background for DIBR 3D images,we propose a hashing authentication scheme for DIBR 3D images,and designed a robust DIBR 3D image hashing scheme based on depth information and local features.In the DIBR process,the depth information of matching feature points is unchanged.According to this characteristics,we can divide the feature points into different groups to generate the final hash value of the image.As the experimental results shown,the proposed hashing scheme is robust to signal distortion attacks such as JPEG compression,adding noise,blurring and gamma correction,moreover,the hashing scheme is sensitive to localized tampering.