Perceptually optimized color image watermarking scheme based on CIEDE2000 color difference equation

In this paper, a wavelet-based color image watermarking scheme is proposed, which is perceptually optimized by a visual model based on the CIEDE2000 color difference equation. The visual model estimates the distortion visibility threshold (DVT) for each wavelet coefficient of each color channel. With the DVT, perceptually significant coefficients in both luminance and chrominance channels can be effectively located for watermark embedding, and the maximum distortion allowed for transparent watermarking can be accurately determined. Through perceptually tuned quantization, binary watermark signals are embedded in perceptually significant coefficients by the method of quantization index modulation. Only a small amount of information, including locations of perceptually significant coefficients and the associated DVT values, are needed for watermark extraction. Experimental results show that the embedded watermark is transparent and quite robust in the face of various attacks such as cropping, median filtering, low-pass filtering, scaling, noise addition, and JPEG compression with high compression ratio.