Isn't it simply hardcore overfitting of network? I haven't read the paper yet but it seems to me that it shows you can use overfitting for output interpolation

Isn't it simply hardcore overfitting of network? I haven't read the paper yet but it seems to me that it shows you can use overfitting for output interpolation with great results. Same should be possibile for example for interpolation of video images instead od pixels.