缺点:
  求得的结果常常与人们的主观视觉效果不一致,这是因为均方误差、峰值信噪比和信号噪声比等都是从整体上反映原始图像和重建图像的差别,并不能反映一幅图像中少数像素点有较大灰度差别和较多像素点有较小差别等各种情况。显然,客观质量评价采用以上各式是对图像中所有的像素点同样对待的,不能全面反映人眼的视觉特性。
  与主观视觉感知相关
  与主观视觉感知相关的方法有: universal image quality index(UIQI),SSIM,VIF,FSI。(上述方法可在reference2中论文中找到)
  ============================================================================================================
  方法的对比
  针对于 image super-resolution 8种方法衡量图片super-resolution之后的质量的对比

  IFC > NQM > WPSNR > MSSSIM > SSIM > UIQI > PSNR >VIF
  好的方法为IFC
  原因:
  1:the IFC metric is designed to evaluate the loss of image information so that it extracts wavelet features with fo cus on high- frequency details rather than low-frequency comp onents. This me tric matches human p erception well as visual p erception is more sensitive to high-frequency details of SR images rather than low-frequency components.
  2:Second, the IFC metric is develop ed based on natural scene statistics using the Gaussian scale mixtures [37] and the BSD200 dataset contains numerous such images.