Texture analysis in watermarking paradigms

Текст статьи не публикуется в открытом доступе в соответствии с политикой журнала.Digital watermarking algorithms have been developed rapidly as a response on the challenges caused by various internet attacks that are distorted the content of the host image and watermark partially or fully. In this paper, the issues of texture analysis with a goal to detect the most suitable image areas for embedding are discussed. The statistical and model-based methods are investigated as a trade-off between the computational cost and quality of the detected areas, where the embedded bits of a watermark could be the most invisible for a human vision. The criteria for detection of such areas based on the textural, contrast, illumination, and color coherence of the host image and watermark are formulated. The experiments show that the statistical methods based on the gradient oriented Local Binary Patterns (LBP) provide better computational time regarding to fractal estimation of textural image areas