Abstract:
Non-negative matrix factorization has become a significant area of research within the last 10 years. After the research paper ”Learning the parts of objects by non-negative matrix factorization” by Daniel D. Lee & H. Sebastian Seung [1] was published, nonnegative matrix factorization was applied to many research areas like text mining and image processing. Since non-negative matrix factorization (NMF) does not provide exact matrix decomposition, iterative methods are being used that depend on many factors like initial conditions and additional constraints that depend on the application requirements. Some of previous researches were aimed to find a unique solution for NMF, on the other hand some of them were based on using NMF with suitable constraints. This thesis studies the watermarking performance a new NMF algorithm that based on fixing on of the resulting matrixes of NMF algorithm. Within this thesis, the multiplicative NMF algorithm introduced by Daniel D. Lee & H. Sebastian Seung [2] was modified ands used for watermarking. The performance of the modified NMF algorithm is analyezd in terms of different parameters with the results of several simulations.
