Algorithms Of Digital Watermarking Based On Singular Value Decomposition In Wavelet Domain

Along with the rapid development in digital,multimedia and network communication technology,the copyright protection of digital media has become increasingly prominent problem which makes digital watermarking technology having been making our close attention because it is one of the most effective solutions to the problem of copyright protection.As the maturation of the wavelet analysis theory and its great advantages in digital image processing,and the singular value decomposition (SVD) matrix transformation has a number of advantages in watermarking algorithm,this paper made intensive studies on wavelet-based watermarking algorithms and SVD-based watermarking algorithms.After deliberation we found that whether watermarking algorithms simply based on various of time-frequency analysises including wavelet analysis or simply SVD-based watermarking algorithms are all have their own weaknesses.The wavelet-based watermarking is robust to common image attacks such as low pass filtering and lossy compression.However,watermarking algorithms simply based on various of time-frequency analysises including wavelet analysis,if not complemented by other improvements,are difficult to resist shear,rotation,zoom and other geometric attacks.Singular values manifest image's inherent algebraic character,which will not severely change when the image suffering a slight perturbation.The singular value decomposition and wavelet transform can be advantages of combining to achieve the compromise of digital watermarking's robustness and invisibility.According to the singular value decomposition is a special matrix transform with good properties,and the wavelet transform is excellent local time-frequence analysis and multianalysis,moreover,block-SVD can offer more larger singular values so as to increase the scope of embedding domain,we proposed a robust digital watermarking algorithm based on block singular value decomposition in wavelet domain:the gray level image is divided into important and unimportant parts,each part is decomposed with blocked-DCT transform after being Arnold scrambled.With the human system (HVS) in the domain of DWT,the original image is decomposed with DWT into several subbands,and block-SVD is applied to transform the subbands.Then embed the important message in the SVD of low freguency suband,and the unimportant message is inserted into the SVD of high frequency subband.Experiments and analysis show that the algorithm have large embedding capacity and strongly robust against the common image processing techniques.Next we performed an adaptive bit-level embedding and adaptive detection for image watermarking algorithm:a wavelet watermarking algorithm for digital image based on SVD and texture features.First we divide the original image into blocks with fixed size,and then we select blocks with abundant image texture features for watermark inserting by analyzing the normalized textural features.Each selected block is decomposed with DWT into several subbands,and SVD is applied to transform the lowest frequency coefficients, then embed watermark adaptively into the SVD of low freguency coefficients.Experimental results show that the embedded digital watermarking algorithm realized the adaptability on embedding and detecting,which is invisible and robust.The watermarking algorithm have large inserting capacity and it is applicable widely on bit-level watermark,which have pragmatic meanings.