Adaptive Image Watermarking Algorithm Based on an Efficient Perceptual Mapping Model

Abstract

Watermark imperceptibility is a significant requirement for keeping watermarked images looking perceptually similar to the original ones. Effective watermark imperceptibility requires the creation of a perceptual model that simulates the human visual system to efficiently hide the watermark in places where the human eye cannot observe it. Current perceptual-based watermarking models use complex computations that are difficult to implement in embedded systems or in real-time applications. In this paper, a low-complexity, integer-based lifting wavelet transform was utilized to create a perceptual mapping model that mainly relies on a new texture mapping model called accumulative lifting difference (ALD). The ALD is combined with a simplified edge detection and luminance masking models to obtain a comprehensive perceptual mapping model that has high-noise tolerance and it is based on low-complexity calculations. The proposed model was 7% faster than the fastest pixel-based compared model with an enhanced average peak signal-to-noise ratio (PSNR) gain of 2.78 dB. In comparison to the largest noise tolerance compared sub-band model, the proposed just noticeable distortion model had a PSNR gain of 1.8 dB and an execution speed that was 90% faster. The perceptual model is utilized in a proposed image watermarking algorithm to determine the maximum watermark embedding intensity that is not visible to the human eye. The experimental results show that the proposed algorithm produced high-quality watermarked images and was robust against different geometric and non-geometric attacks. In addition to its usage in watermarking, the new perceptual model can be used in various image processing and real-time applications.