A Digital Watermarking Algorithm Based on Non-Negative Matrix Factorization and Orthogonal Projection Transformation

This paper presents a watermarking algorithm based on non-negative matrix factorization (NMF) and orthogonal projection transformation. Firstly, the base matrix of an image is constructed by NMF, which is the part-based image representation. Then, the base matrix is converted perpendicularly and it is regarded as a key to extract the watermark later. The watermark is embedded into the coefficient matrix, which is made by mapping the image to the orthogonal base matrix. Finally the image is reconstructed by inverse transform of the matrix. The presented algorithm demonstrates an obvious advantage in precision of reconstruction because the basic NMF method is improved by above steps. It has been applied to the digital watermarking system and the result shows better robustness and practicability than that reported in reference 4.