Research On Screen-cam Robust Watermarking Model And Algorithm For Remote Sensing Images

Remote sensing image data is a national strategic information resource,which plays an important role in national defense construction,engineering planning,and disaster monitoring.In order to protect remote sensing image data,encryption technology,access control technology,and digital watermarking technology have been widely used.However,with the ubiquity of computerized work and smartphones,new security issues arise.Using mobile phones to capture the content displayed on the computer screen has made the leakage of remote sensing images much easier and more frequent.The encryption and access control technologies cannot prevent leakers from taking pictures of plaintext data.Furthermore,traditional digital watermarking technology is not applicable in this scenario.Therefore,effectively protecting remote sensing image data and holding accountability for sneak shots need to be solved as an urgent problem.The screen-cam robust watermarking technology for remote sensing image data embeds an imperceptible watermark in the data.After the camera device captures the watermarked remote sensing image displayed on the screen,the watermark can still be extracted from the recaptured photos.In addition,this watermarking scheme is also robust against common image processing attacks.Therefore,the screen-cam robust watermarking technology for remote sensing image data provides an effective solution for protecting data and holding accountability for leakage.Furthermore,we not only have to trace the leaker after the leakage but also to prevent the occurrence of leakage.Encryption technology is a typical security technology to prevent data leakage,which can also prevent people who do not have the key from using plaintext data after encrypting remote sensing images.Hence,encryption technology can reduce the possibility of data leakage.To fully protect remote sensing image data,we need to combine encryption technology and screen-cam robust watermarking technology and achieve key management based on watermarking technology.This paper studies the theory,schemes and algorithms of screen-cam robust watermarking for remote sensing images.The main research works and results are as follows:(1)We first analyze each sub-process of the screen-cam in theory.Then,the signal distortions caused by the screen-cam attack are summarized into five categories: linear distortion,gamma adjustment,geometric distortion,noise attack and low-pass filter attack.After that,we perform quantitative analysis of the variation characteristics of remote sensing images during the screen-cam process through experiments.Based on this,we explored the possibility of spatial domain or transform domain as watermark carrier,local feature operators as watermark synchronization intermediary.(2)We propose a feature-based screen-cam robust watermarking model to achieve robustness against both screen-cam attack and common image processing attacks.At the algorithm implementation side,a local feature square region construction method based on Gaussian preprocessing,Harris-Laplace operator,and SURF(Speeded-up Robust Features)feature orientation operator is designed.This feature region is stable to screen-cam attack,and invariant to orientation and scale,effectively achieving watermark synchronization against screen-cam attack.We employ a non-rotating embedding and a preprocessing method to improve the DFT(Discrete Fourier Transform)-based watermarking embedding algorithm.At the same time,a watermark extraction algorithm based on local statistical features is proposed.Experiment results show that the algorithm will not affect the use of data,and has good robustness against both screen-cam attack and common image processing attacks.(3)We present a pattern-based screen-cam robust and blind watermarking model to achieve blind watermark detection and automatic extraction.At the algorithm implementation side,a watermark pattern construction and embedding method is designed.At the watermark extraction side,a synchronization response index is proposed.Using this index,we can estimate the approximate positions of synchronization watermark from multi-scales of the screen-cam photos,and calculate the precise coordinates from the local areas of the approximate positions.Thereby,we can achieve blind watermark synchronization from a remote sensing image with perspective distortion.Experiment results show that the algorithm will not affect the use of data,and has high robustness against both screen-cam attack and common image processing attacks.(4)We propose a joint encryption and screen-cam robust watermarking model to take advantage of both encryption and watermarking technology and achieve key management.At the algorithm implementation side,a dual watermark embedding algorithm is designed,which hides one watermark in the image variations caused by embedding another watermark.Based on the designed method,we can guarantee both watermarks and achieve screen-cam robust watermarking for plaintext data.At the same time,a joint mechanism based on odd-even segment encryption and odd-even quantization-based watermarking is designed.Based on this joint mechanism,the perceptibility of watermark can be enhanced after encryption,so as to screen-cam robust watermarking for ciphertext data.Experiment results show that the algorithm is security,and will not affect the use of data.Also,the ciphertext data and plaintext data both have high robustness against screen-cam attack.Besides,the plaintext data has good robustness against common image processing attacks.