regarding autoencoder

[viraj moil]

Hii, I am new keras user. I am using denoising autoencoder for denoise the image in unsupervised way.However I have some queries that I need clarification.

How the denoising autoencoder know it is noise or signal and ow it separates signal from noise

Secondly what actually the latent space content.can anybody please suggest a better understandable solution to my queries.