Abstract
Digital watermarking provides copyright protection and proof of ownership by inserting watermark metadata as owner's identity in digital documents to prevent authenticity and copyright violations. The paper introduces a new hybrid image watermarking scheme by attaching multiple copies of watermarks in carrier image. The new scheme utilizes the advantages of DWT, DFT, DCT, and SVD transformations to offer stable resistance in protecting watermark contents from various external attacks. The proposed scheme uses Haar wavelet, Fourier, Onion Peel Decomposition, DCT, zigzag ordering, and SVD transformations to decompose the carrier image into four levels to maintain imperceptibility in the watermarked images. The algorithm attaches replicas of watermark frequency blocks in all frequency components of host image to provide better robustness against external deprivations in watermarked images. The proposed algorithm also provides the increased probability of extracting at least one undamaged replica of watermark even when other frequencies are damaged by external attacks. The improved experimental results of the proposed scheme in terms of visual analysis and quantitative metrics on different images with different experimental setup demarcate that the proposed watermarking scheme provides stable performance in generating better watermarked images. It is experimentally found that the new scheme produces high quality watermarked images with an average of 7.62% lesser Mean Absolute Error (MAE) and increased Peak Signal to Noise Ratio (PSNR), Mean Structural Similarity Index Measure (MSSIM), and Feature Similarity Index Measure (FSIM) of 5.02 %, 4.37 %, and 2.37 %, respectively, than the next best algorithms when simulated with 20 sets of watermark and cover images. The watermark images extracted by the proposed algorithm from extremely distorted watermarked images are with better visual and objective values than other methods used in the comparative study. Simulation analysis on 20 sets of watermark and cover images with 30 types of potential attacks reveals that the extracted watermark images through the proposed scheme produces an average of 5.62%, 6.37%, and 5.75% improved Pearson Correlation Coefficients (PCC), Number of Changing Pixel Rate (NPCR), and the Unified Averaged Changed Intensity (UACI) values, respectively, than the next best algorithms used in the comparative study.