Research On Fragile Watermarking Schemes For Image Content Authentication And Self-recovery

The popularity of mobile terminal and the development of Internet make image acquisition and transmission become more and more convenient.At the same time,more abundant image processing methods and advanced image editing software make it increasingly easy for ordinary people to modify and retouch the digital images than ever before.However,this also brings about a serious problem that a lot of tampered or forged images spread on the social platforms due to the convenience of Internet.These images not only strike against people's traditional cognition which is "to see is to believe",but also affect people's judgement to the public trust of social media.Therefore,it becomes particularly important to research the digital image content forensic technology.To solve the above issue,this thesis starts with digital watermarking-based authentication theory and studies the fragile watermarking method for image content authentication and self-recovery,which can further restore the forged areas after tamper location.The principal research work is as follows:(1)To resolve the tampering coincidence problem in mapping block-based fragile watermarking method,and improve the tamper location accuracy and the quality of restored images,this thesis designs a fragile watermarking method with multilayer tamper location and recovery mechanism.Based on the repeated embedding strategy,the average value of an image block is served as the first recovery data and inserted into a mapping block obtained by the first mapping sequence,while the difference value between the image block and its first mapping block is served as the backup recovery data and inserted into another mapping block obtained by the second mapping sequence.In the detection side,the three-level tamper location mechanism is adopted,which can greatly improve the detection accuracy for tampered regions.To restore the tampered blocks,the first recovery data in the first mapping block is preferentially selected according to the embedding rules.If a block and its first mapping block are both determined as forged blocks,the backup recovery data in the second mapping block is extracted.The tampered block is reconstructed by the valid recovery data and iterative restoration method.Simulation experiments indicate that the designed fragile watermarking method can provide multiple opportunities for the restoration of a detected forged block.The reconstruction failure problem caused by tampering coincidence can be reduced significantly.(2)To make the recoverable fragile watermarking method can allocate different amounts of recovery data for different blocks and images with different texture characteristics,a self-recovery fragile watermarking with alterable watermark capacity is designed,which is based on the discussion of Singular Value Decomposition(SVD).By analyzing the relationship between the product of left and right singular vectors of an image block after SVD transform and the textural features of the block,a new block classification strategy is introduced.The blocks are segmented into two categories according to its roughness:smooth blocks and texture blocks.For different blocks,alterable-length recovery watermark is obtained by different recovery data generation methods.In other words,the watermark capacity is alterable for different images.Experimental results indicate that the designed fragile watermarking method takes the texture characteristics of image blocks into full consideration at image reconstruction stage.The restored images have similar image quality as the original watermarked images.