Abstract

The rapid growth of digital multimedia data and the increasing use of the World Wide Web have triggered the need for the protection of multimedia contents (images, 3D graphics, audio, video). Digital watermarking has been proposed as an effective solution to the multimedia protection. The great challenge, however, is to devise efficient and robust watermarking schemes to tackle the fundamental tradeoff between robustness, data payload, and imperceptibility. This thesis is devoted to two robust watermarking schemes that we have developed for copyright protection. The first watermarking scheme uses discrete wavelet transform (DWT) and nonnegative matrix factorization (NMF). The core idea consists of decomposing an image into four wavelet sub-bands and then applying NMF to the blocks of each sub-band, followed by an eigen-decomposition distortion step. The second watermarking scheme uses multiple parameter discrete fractional Fourier (MPDFRF) transform and DWT. The original cover image is decomposed into four wavelet sub-bands, followed by a segmentation of each sub-band into blocks. Then the MPDFRF transform is applied to each block. The rest of the thesis is devoted to 3D mesh compression and fingerprinting. We propose a mesh compression technique using the mesh umbrella matrix. The key idea is to apply eigendecomposition to the umbrella matrix, and then retain the eigenvalues and eigenvectors which contain the low frequency coefficients. We also propose a mesh fingerprinting method by partitioning a 3D mesh into sub-meshes to produce the fingerprint vectors for all the sub-meshes