Autoencoder result color distortion

I have posted a question about the normalized image seems to be a bit “off”

Relating to the problem, the output of the image compared to the input seems rather color distorted.

I just one someone to point out if this is a problem that convT faces in definition, or if it’s me that’s doing something wrong?

The papers that I had seen on GANs, and Autoencoders never seemed to have such problems.