Blind Image Watermarking Based on Local Analysis of Gradients

Digital image watermarking has been emerged as a basic method for copyright protection and authenticity of the owner. This paper proposes a novel and adaptive blind watermarking method using local analysis of gradients in an image block. The method partitions the image into non-overlapping blocks. The embedding is performed in the transfer domain of each image block. Two transform coefficients are modified using a variable strength factor. The value of strength factor depends on the local complexity of the image. This value is adaptively obtained from the mean gradient of each block and the DC element of the DCT coefficients of the block. Experimental results under different attacks such as JPEG compression, additive noises and filters show improvements in robustness of our method as compared to some state of the art methods in the literature.