Nearest-neighbor and Bilinear Resampling Factor Estimation to Detect Blockiness or Blurriness of an Image

摘要

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. In this paper, we outline novel nearest-neighbor and bilinear interpolation detection algorithms that are de-signed to estimate rational resampling factors (above 1 x) in both the vertical and horizontal dimensions. The robustness of these algorithms to several common post-processing algorithms is also evaluated.