The Color Image Watermarking Algorithm Based On Energy Loss And Data Decomposition

In the latest several years, the sharp development of digital techniques andcomputer network gives the application of digital multimedia (digital images, digitalaudio, digital video and3-D computer images) a stupendous development and bringalong with our work and life great convenience. But just because of this, digitalmultimedia is very easy to be copied, juggled, illegally diffused and purposiveattacked. So, the problem of copy right protection to digital multimedia is being givenmore and more attention and digital watermarking technique is so far the mosteffective digital product copyright protection technique. In this paper, we primarilyinvestigate and discuss the color image digital watermarking in this field.For the traditional color images, the researchers often embed the watermarkinformation into the luminance component or a single color component. Suchalgorithms not only do not consider the correlation of the three-component R, G, andB, but also it will make the robustness and visibility of the watermark informationweek. The existing watermarking algorithms, based on the data decomposition, havebeen able to make the watermark information spread to the three components R, G,and B, while the processing of vectorization and matrixing do not fully take intoaccount the correlation of the three components and will make the energy lost.This paper proposes a color image watermarking algorithm, which combines theTucker decomposition and HOSVD decomposition. First, do DCT transform for thethree-component of R, G and B and take the DC component to compose the DCmatrix: DCR, DCG and DCB; second, construct the3-level tensor with the three DCmatrixes and get the core tensor from Tucker decomposition; At this step,watermarking algorithms in the existing data decomposition, often embed thewatermark information into the quantifying or matrixing core tensor, which led to theenergy loss of the R, G and B components. third, do the SOSVD transform for thecore tensor and obtain three indicators of the core tensor; fourth, get the singularvalues of the three indicators; last, according to the watermark information to modify the singular values to embed the watermark information.The experimental results show that, compared with the conventional color imagewatermarking algorithm and the existing high-dimensional data decompositionalgorithm in the conventional image processing (noise, filtering, zoom) and shearattacks, the color image watermarking algorithm has better robustness and invisibility.