Multiple embedding of watermarks into a spatial-frequency domain of images based on a genetic algorithm

The widespread use of digital content makes the task of protecting author’s and owner’s rights increasingly important, in particular with regard to digital images. Digital watermarking technology is an effective tool that solves many problems associated with proving authorship of images, verifying authenticity, and tracking illegal copying. An effective watermarking algorithm requires achieving high levels of imperceptibility and robustness, which is a difficult task, since improving one of these indicators usually leads to a deterioration in the other. This study proposes a new watermarking algorithm in a hybrid spatial-frequency image domain based on multiple embedding and metaheuristic optimization. Watermark embedding is done by changing a block of image pixels according to some change matrix, which is selected adaptively for each block using a genetic algorithm. During the extraction stage, a value of each watermark bit is determined using all embedded copies. Neither an original image nor an original watermark is required for data extraction. Experimental results show that the proposed algorithm is highly imperceptible and resistant to a number of image processing attacks.