Research On Robust Reversible Algorithms For Digital Images

Reversible watermarking,as a special watermarking technology,requires the extraction side to recover the original medium without distortion after extracting the watermark correctly.This characteristic,i.e.,reversibility,is critical for some special applications with high fidelity requirements like medicine,military,law scenes,and so on.However,the cover may be attacked in the transmission,requiring that the reversible watermarking is resistant to some non-malicious attacks,such as JPEG compression and noise.Therefore,robust reversible watermarking(RRW)is then proposed to meet the requirement that the watermark extraction and original image recovery can be achieved in a losslessly environment,while the watermark can still be correctly extracted after attacked.Most of the existing RRW algorithms are designed for BMP images,and few algorithms are proposed for JPEG images.Also,the performance of these algorithms,including robustness and visual quality,is far from satisfactory.Therefore,in this paper,by summarizing existing algorithms,three RRW algorithms are proposed as follows to improve the embedding performance.(1)A RRW algorithm based on redundant histogram shifting(RHS)with high fidelity is proposed.A BMP image is first divided into non-overlapping blocks,and a high pass filter is applied to each block to generate a histogram which is a Laplacian-like distribution.Then,the watermark is embedded into blocks by shifting the generated histogram.Specifically,following the histogram shifting rule,the embedding distortion is minimized based on a new modification mechanism.That is,the histogram shifting is conducted by modifying each pixel in the block with a specific rule.Moreover,for blind data extraction and image recovery,a new strategy for determining the parameters used in histogram shifting is also proposed.Experiments verify that the proposed algorithm has better visual quality while ensuring robustness compared with the previous algorithm.(2)A RRW algorithm for JPEG images based on pairwise prediction error expansion(Pairwise PEE)is proposed.Firstly,the diamond prediction is performed on direct coefficients of the quantized discrete cosine transform(DCT).Secondly,a two-dimensional prediction error histogram is generated by scanning adjacent DCT blocks as a pair.Finally,the watermark is embedded by modifying the prediction error histogram.In the proposed algorithm,the local correlation of neighboring direct coefficients is effectively exploited,thus reducing the embedding distortion.In addition,experiments demonstrate that the quality of images after embedding is effectively improved and the robustness to JPEG recompression is achieved in our algorithm.(3)A RHS-based RRW algorithm for JPEG images is proposed.To overcome the low robustness of existing algorithms,RHS is employed to embed the watermark into quantized DCT low-frequency coefficients.Specifically,DCT blocks are firstly divided into four groups by rows.Then,the sum of low-frequency coefficients in the same frequency band and the same group of DCT blocks is calculated to generate a distribution histogram.Finally,the watermark information is embedded by shifting the histogram.In addition,by employing high-frequency quantized DCT coefficients to predict the image complexity,the watermark can be adaptively embedded.Experimental results show that the proposed method can effectively resist JPEG recompression while ensuring the visual effect.