We are comparing against the method described in:

Y. Pritch, E. Kav-Venaki, and S. Peleg,
Shift-Map Image Editing,
ICCV'09, Kyoto, Sept. 2009, pp. 151-158

We contacted the authors of the shift-map publication, who provided the results below to us. They didn't have their original shift-map implementation anymore; the results below were generated using an implementation written by a student in their group, however. The implementation wasn't able to run on the full resolution images, so all images below are resized to quarter resolution. In addition it hung on a few specific img+mask instances. Hence some images from the other comparisons are missing here and some others use a slightly modified mask.

For the reasons stated above we didn't include the Shift-Map results in the user study and instead provide a qualitative comparison against our method here. Our method was applied to same input images as Shift-Map (downsampled and modified masks).