Image processing algorithms always affect the quality of the image to which they are applied. Traditionally, the effects of these algorithms are determined using broad image quality measurements such as PSNR, SSIM, VIF and MSE. These quantitative measures, however, do not provide a functional assessment of the image processing algorithms. In this paper, we propose and introduce the use of facial detection as such a functional measurement of image quality after performing image processing transformations on the image. To assess this, we downsampled images containing a combined 953 faces, then upsampled the images using five different image processing operations including interpolation and filtering with different parameters. This process resulted in the misdetection of 12-20 images (and a 14-21% increase in error rate) in comparison to the original set, illustrating the deleterious effect of downsampling and subsequent upsampling on the images.
Steven Simske, Dalong Li, Darryl Greig, "Use of Face Detection to Qualify Image Processing Algorithms" in Proc. IS&T Int'l Conf. on Digital Printing Technologies and Digital Fabrication (NIP28), 2012, pp 234 - 235, https://doi.org/10.2352/ISSN.2169-4451.2012.28.1.art00068_1