Entroopial Põhinev Vastupidav Vesimärgi Algoritm
Abstract
Tänu aina kasvavale multimeedia andmeedastus mahtudele Internetis, on esile kerkinud mured turvalisusest ja piraatlusest. Digitaalse meedia paljundamise ja muutmise maht on loonud vajaduse digitaalse meedia vesimärgistamise järgi. Selles töös on tutvustatud vastupidavaid vesimärkide lisamise algoritme, mis lisavad vesimärgid madala entroopiaga pildi osadesse. Välja pakutud algoritmides jagatakse algne pilt blokkidesse ning arvutatakse iga bloki entroopia. Kõikide blokkide keskmine entroopia väärtus valitakse künniseks, mille järgi otsustatakse, millistesse blokkidesse vesimärk lisada. Kõik blokid, mille entroopia on väiksem kui künnis, viiakse signaali sageduse kujule kasutades Discrete Wavelet Transform algoritmi. Madala sagedusega sagedusvahemikule rakendatakse Chirp Z-Transform algoritmi ja saadud tulemusele LU-dekompositsiooni või QR-dekompositsiooni. Singular Value Decomposition meetodi rakendamisel diagonaalmaatriksile, mis saadi eelmisest sammust, saadakse iga bloki vastav väärtus. Vesimärk lisatakse pildile, liites iga bloki arvutatud väärtusele vesimärgi Singular Value Decomposition meetodi tulemused. Kirjeldatud algoritme testiti ning võrreldi teiste tavapärast ning uudsete vesimärkide lisamise tehnoloogiatega. Kvantitatiivsed ja kvalitatiivsed eksperimendid näitavad, et välja pakutud meetodid on tajumatud ning vastupidavad signaali töötlemise rünnakutele.
With growth of digital media distributed over the Internet, concerns about security and piracy have emerged. The amount of digital media reproduction and tampering has brought a need for content watermarking. In this work, multiple robust watermarking algorithms are introduced. They embed watermark image into singular values of host image’s blocks with low entropy values. In proposed algorithms, host image is divided into blocks, and the entropy of each block is calculated. The average of all entropies indicates the chosen threshold value for selecting the blocks in which watermark image should be embedded. All blocks with entropy lower than the calculated threshold are decomposed into frequency subbands using discrete wavelet transform (DWT). Subsequently chirp z-transform (CZT) is applied to the low-frequency subband followed by an appropriate matrix decomposition such as lower and upper decomposition (LUD) or orthogonal-triangular decomposition (QR decomposition). By applying singular value decomposition (SVD) to diagonal matrices obtained by the aforementioned matrix decompositions, the singular values of each block are calculated. Watermark image is embedded by adding singular values of the watermark image to singular values of the low entropy blocks. Proposed algorithms are tested on many host and watermark images, and they are compared with conventional and other state-of-the-art watermarking techniques. The quantitative and qualitative experimental results are indicating that the proposed algorithms are imperceptible and robust against many signal processing attacks.
With growth of digital media distributed over the Internet, concerns about security and piracy have emerged. The amount of digital media reproduction and tampering has brought a need for content watermarking. In this work, multiple robust watermarking algorithms are introduced. They embed watermark image into singular values of host image’s blocks with low entropy values. In proposed algorithms, host image is divided into blocks, and the entropy of each block is calculated. The average of all entropies indicates the chosen threshold value for selecting the blocks in which watermark image should be embedded. All blocks with entropy lower than the calculated threshold are decomposed into frequency subbands using discrete wavelet transform (DWT). Subsequently chirp z-transform (CZT) is applied to the low-frequency subband followed by an appropriate matrix decomposition such as lower and upper decomposition (LUD) or orthogonal-triangular decomposition (QR decomposition). By applying singular value decomposition (SVD) to diagonal matrices obtained by the aforementioned matrix decompositions, the singular values of each block are calculated. Watermark image is embedded by adding singular values of the watermark image to singular values of the low entropy blocks. Proposed algorithms are tested on many host and watermark images, and they are compared with conventional and other state-of-the-art watermarking techniques. The quantitative and qualitative experimental results are indicating that the proposed algorithms are imperceptible and robust against many signal processing attacks.