基于显著性的图像质量评价方法研究

Evaluation method study of image quality based on significance

随着视觉生理学和视觉心理学的发展,对HVS研究的深入,出现了很多基于视觉、感知的全参考质量评价方法,又由于失真大概可以分为两种:随机噪声和模糊,故提出了一种基于显著性的评价方法 (SIQA),利用残差图像(参考图像与失真图像之差)局部的统计均值及方差来反映噪声的强度,利用边缘向量来反映模糊的程度,在关注区域权值增大显然符合人类视觉系统。

With the development of visual physiology and psychology, and the thorough research of HVS, there are a lot of vision based, full reference quality evaluation method of perception. The distortion can be roughly divided into two kinds： random noise and fuzzy, so this paper presents an evaluation of significant method based on SIQA using the residual images （the reference image and the distortion of image difference） statistical mean and variance of partial to reflect the intensity of noise, using the edge vector to reflect the fuzzy degree. The weight increased in the region of interest obviously consistent with the human visual system.