I conduct Super Resolution task, should I normalize the target and input?

Question

I run the LapSRN model this days. The input and target of this model are both images, and I split the RGB channels to train different channels.
And the mean and std of