基于频率、空间和方向的图像质量评估

Image Quality Measurement Based on Scale,Space and Orientation

将空间域图像质量评价方法结构相似度SSIM推广到HWD变换域,结合人眼视觉倾斜效应和粒子群优化算法,提出一种新的图像质量评价测度。将SSIM直接用于各HWD分解频带,用频带相关性图加权各频带的结构相似度得到局部质量,然后对不同方向的局部质量求加权和得到整幅图像的结构相似度。实验结果表明,该测度与主观感知有较好的一致性,能准确地反映人眼对图像的视觉感知。

In this paper, a metric that can measure image quality across scale, space and orientation （SSO- IQM） is proposed based on the fact that human visual system is highly tuned to the stimuli which are ＂narrow-band＂ in frequency, space and orientation. The newly proposed metric compares the structural similarity between the refer- ence and the distorted image in the HWD space, which is then weighted by the correlation map to obtain the local quality measure for each subband. To measure image quality across orientation, the weighting factors of the subba- nds are assigned different values experimentally. Furthermore, extensive experiments indicate that the proposed met- ric SSO-IQM is more consistent with the subjective evaluation than all the other competing methods evaluated.