Abstract: Reduced-Reference（RR） Image Quality Assessment（IQA） aims to predict the visual quality of distorted images with only partial information concerning the reference images. In this paper, a RR IQA based on digital watermarking contourlet domain is proposed. Firstly, the low-frequency sub-band of contourlet transform is selected as the watermark embedding region based on the theory of Human Visual System （HVS）, so semi-fragile digital watermarking is constructed. Secondly, the Structural Similarity（SSIM） value between embedded watermark image and original image is used to design an adaptive watermark embedding system, which ensures the invisibility of watermark. Finally the image quality metric is gained by the restoration rate of watermark of distorted image. Empirical studies are carried out upon the LIVE images database II and TID2008 images database against the subjective IQA score and demonstrate that the proposed framework has favorable consistency with subjective perception values and the objective assessment results can well reflect the visual quality of JPEG compressed images.

Key words: contourlet transform, digital watermarking, Image Quality Assessment（IQA）, Structural Similarity（SSIM）

摘要： 半参考图像质量评价方法是一种利用原始图像的部分信息对失真图像进行质量预测的方法，提出了一种基于轮廓波变换数字水印的JPEG图像压缩半参考质量评价方法。首先参考人类视觉系统（Human Visual System，HVS）的思想，选取轮廓波变换的低频区域作为水印嵌入区域，从而生成半脆弱数字水印。然后根据已嵌入水印图像与原始图像之间的结构相似度（Structural Similarity，SSIM）的值，设计自适应水印嵌入系统，以保证水印的不可见性。最后分别在LIVE图像数据库2和TID2008图像数据库中，根据已嵌入水印图像进行数据库重建，并测试该算法的性能。实验结果对比显示，该算法与其他算法相比较，具有较好的准确性、单调性以及一致性，能够较好地反应JPEG失真图像的质量。

关键词: 轮廓波变换, 数字水印, 图像质量评价, 结构相似度（SSIM）