CoMoFoD - Image Database for Copy-Move Forgery Detection

CoMoFoD database

CoMoFoD database has 260 image sets, 200 images in small image category (512x512), and 60 images in large image category (3000x2000). In both categories, following transformations are applied:

  1. translation - a copied region is only translated to the new location without performing any transformation,
  2. rotation - a copied region is rotated and translated to the new location,
  3. scaling - a copied region is scaled and translated to the new location,
  4. distortion - a copied region is distorted and translated to the new location,
  5. combination - two or more transformation are applied on a copied region before moving it to the new location.

Every image set consist of:

Original image Mask Binary mask Forged image

 

Postprocessing methods applied on all original and forged images are:

Image naming

Images are named as follows: N1_M1, where N1 is three (small image category) or two (large image category) digits equal to image number in category, and M1 is one of four different marks:

  1. "F" - forged image,
  2. "B" - binary mask (black and white mask),
  3. "M" - mask (colored mask),
  4. "O" - original image.

Additional information about transformations applied on forged images can be seen in "readme.txt" files.
Postprocessed images are named as follows: N1_M1_M2, where additional mark M2 points postprocessing methods applied on image:

  1. "JC" - JPEG compression, quality factor = [20, 30, 40, 50, 60, 70, 80, 90, 100],
  2. "IB" - image blurring, μ = 0, σ2 = [0.009, 0.005, 0.0005]),
  3. "NA" - noise adding, averaging filter = [3x3, 5x5, 7x7],
  4. "BC" - brightness change, (lower bound, upper bound) = [(0.01, 0.95), (0.01, 0.9), (0.01, 0.8)],
  5. "CR" - color reduction, intensity levels per each color channel = [32, 64, 128],
  6. "CA" - contrast adjustments, (lower bound, upper bound) = [(0.01, 0.95), (0.01, 0.9), (0.01, 0.8)].
Usage of different parameters in postprocessing methods was marked with numbers at the end of image in such way that lower number correspond to lower used parameter value.