文章基本信息

标题：An Experimental Study into Objective Quality Assessment of Watermarked Images
本地全文：下载
作者：Dr. Anurag Mishra ; Mr. Aruna Jain ; Mr. Manish Narwaria 等
期刊名称：International Journal of Image Processing (IJIP)
电子版ISSN：1985-2304
出版年度：2011
卷号：5
期号：2
页码：199-219
出版社：Computer Science Journals
摘要：In this paper, we study the quality assessment of watermarked and attacked images using extensive experiments and related analysis. The process of watermarking usually leads to loss of visual quality and therefore it is crucial to estimate the extent of quality degradation and its perceived impact. To this end, we have analyzed the performance of 4 image quality assessment (IQA) metrics - Structural Similarity Index (SSIM), Singular Value Decomposition Metric (M-SVD) and Image Quality Score (IQS) and PSNR on watermarked and attacked images. The watermarked images are obtained by using three different schemes viz., (1) DCT based random number sequence watermarking, (2) DWT based random number sequence watermarking and (3) RBF Neural Network based watermarking. The signed images are attacked by using five different image processing operations. We observe that the metrics behave identically in case of all the three watermarking schemes. An important conclusion of our study is that PSNR is not a suitable metric for IQA as it does not correlate well with the human visual system's (HVS) perception. It is also found that the M-SVD scatters significantly after embedding the watermark and after attacks as compared to SSIM and IQS. Therefore, it is a less effective quality assessment metric for watermarked and attacked images. In contrast to PSNR and M-SVD, SSIM and IQS exhibit more stable and consistent performance. Their comparison further reveals that except for the case of counterclockwise rotation, IQS relatively scatters less for all other four attacks used in this work. It is concluded that IQS is comparatively more suitable for quality assessment of signed and attacked images.
关键词：Digital Image Watermarking ; Image Quality Assessment; PSNR; M-SVD; SSIM; Image Quality Score