Research On Selection Encryption And Lossless Reconstruction In JPEG Images

With the popularization of 5G,emerging technologies such as cloud computing and cloud storage are devolping rapidly.Various types of data such as documents and audio can be transmitted swiftly through the communication networks.However,while convenience has been brought by communication technology,the risk of leakage of privacy contained in large amounts of data has been greatly increased.Personal private information,confidential business messages and even important government secrets can not be guaranteed secure and reliable when transmitted over the network.These data may be eavesdropped,captured or tampered with by malicious attackers.Therefore,it is urgent to research on protecting the privacy data.As a type of data like documents and audio files,images also contain privacy information.Since JPEG standard is one of the most popular image format standard,it is important to protect the privacy in JPEG images.The area contains privacy information,namely region of interest(ROI),is usually just part of the whole image.In order to save the unnecessary cost of encryption,the encryption is only applied in ROI,which is called selection encryption.In the sake of storage,the encryption algorithm should be reversible,that is,the original image can be lossless reconstructed from the encrypted image.Moreover,the file size of the encrypted image is another considerable problem in JPEG privacy protection.Based on this,two JPEG-based selection encryption and lossless reconstruction(SELR)schemes are proposed in this thesis.The main work and innovation points of this thesis are listed as follows.1.Propose an image-blurring based SELR in JPEG imageAfter the ROI is extracted by methods such as face detection,the first several quantized DCT coefficients are modified to their average to blur the ROI.Then the reconstruction information is embeded into the image.Variable parameters are introduced in this scheme.Optimal encrypted results can be achieved by selecting appropriate parameters for different practical requirements.Moreover,multi-ROI overlapping problem is discussed and solved.Under the requirement of PSNR and file size increment,experiments were performed to find out how paremeters influence the results and recommend parameters were found.Experimental results demonstrate that the proposed scheme outperforms the existing ones on privacy protection and file size with proper parameters.2.Propose an image-transformation based SELR in JPEG imageThe image-blurring scheme above has shortcomings of uncontrollable inflexible and unattractive visual effects.To improve the visual effects of protected ROI,imagetransformation based SELR in JPEG image is proposed,in which ROI is visually transformed into a specified image,i.e.the target image.A block-based image transformation algorithm is proposed according to the features of JPEG standard.Meanwhile,the distortion of the area outside ROI,i.e.background is minimized.Experimental reults demostrate that the proposed scheme realizes good image transformation effect with various target images,and the backgound has nearly no change.The proposed scheme fills in the blank in JPEG image transformation.