基于小波域数字水印的图像质量评价方法

为了进一步提高部分参考型图像质量评价方法的准确性和水印隐藏性,提出了一种基于小波域数字水印的图像质量评价方法.首先结合人类视觉系统的特性检测原始图像的纹理丰富区域,获得水印嵌入指示图;然后将原始图像进行3层小波分解,根据水印嵌入指示图,在每个小波系数块内嵌入自定义水印,运用量化参数自动调节系统确保数字水印的隐藏性;最后通过失真图像的水印复原率实现对图像的客观质量评价.与传统的图像客观质量评价方法(PSNR)和部分参考型图像质量评价方法(RRIQA)的对比实验表明,文中提出的方法在预测精确性上分别高出8.5%和4.4%,单调性上分别高出5.4%和2.2%,一致性上分别高出5.7%和3.4%.

Image Quality Assessment Based on Digital Watermarking in Wavelet Domain

A typical kind of reduced-reference image quality assessment is to compare the watermark embedded in the original images with the one extracted in the distorted images. Firstly, the proposed metric acquires the watermark embedded map based on the characteristics of human visual system. Then, the 3-level discrete wavelet decomposition is implemented, and the digital watermark is embedded into each wavelet coefficient block according to the watermark embedded map. Finally, the image objective quality is predicted by the watermark recovered rate from the distortion image. The adaptive adjustments of quantization parameter are applied to ensure the watermark’s invisibility. The comparing experiments show that the prediction accuracy, monotonicity and consistency of the proposed metric are respectively 4.4%, 2.2% and 3.4% higher than that ones of the PSNR and the Wang’s metric.