New Blind Image Moments Watermarking Method Based on Evolutionary Optimization of Embedding Parameters

Abstract

In image watermarking, the locations where the watermark is embedded in the frequency domain and the embedding strength influence the overall performance of the blind watermarking procedure. The present paper aims to propose a new blind watermarking method based on the dither modulation by developing an automatic selection of the optimum embedding parameters that guarantee high-quality watermarked images and low bit error rates during the extraction process. The proposed automatic search method for the best discrete moments subsets is based on an evolutionary algorithm and adopts a specific coding strategy with a group of genes representing the embedding positions. The second part of the chromosome is reserved for the embedding strength coding, followed by the application of different evolutionary operators on the evolution pool. Our study explores the impact of maximum generation, population size, and cutting-point positions with different crossover and mutation rates. The performances under different attack conditions are evaluated, and a comparative study is established with other conventional selection methods and other discrete transforms. Results show that our proposed optimization algorithm baptized EWIMps achieves the best trade-off between robustness and imperceptibility with a peak signal-to-noise ratio varying from 28.33 dB to 59.87 dB and a normalized cross-correlation value from 0.707 to 1.