Abstract: In digital publishing, a low-resolution image is highly undesirable. Inexperienced users often try to include low-resolution images from the Internet or digital cameras in documents they are composing. Current preflight tools are able to single them out, but what if those low-resolution images have been interpolated? They may have a sufficient resolution, but their quality has been compromised, especially images interpolated by nearest-neighbor (which includes pixel replication) and bilinear interpolation. The interpolated images often display blocky artifacts, blurry artifacts, or loss of texture. We outline novel nearest-neighbor and bilinear interpolation detection algorithms that are designed to estimate rational resampling factors (above 1×) in both the vertical and horizontal dimensions. The robustness of these algorithms to several common postprocessing algorithms is also evaluated.

@article{Suwendi:2008aa,
  author       = {Ariawan Suwendi and Jan P. Allebach},
  url          = {http://link.aip.org/link/?JEI/17/023005/1},
  number       = {2},
  pages        = {023005},
  volume       = {17},
  title        = {Nearest-neighbor and bilinear resampling factor estimation to detect blockiness or blurriness of an image},
  year         = {2008},
  keywords     = {Image Forensics; Tamper Detection},
  journal      = {Journal of Electronic Imaging},
}