I've just made available the test errors for our method. This is the
data used to generate the "Our method" CDF in our paper[1]. The errors
are available as a text file on the project webpage. This should make
comparisons easier.
http://people.csail.mit.edu/vladb/photoadjust/
[1] "Learning Photographic Global Tonal Adjustment with a Database of
Input / Output Image Pairs"