Essentially there are only 2 kinds of image.....lossy and loss-less.
If an image is going to have any post-drawing modification...ie alterations-resaving or resizing, etc it NEEDS to be 'loss-less' before such.
Once any work and/or rescaling is done it can be 'lossy' [compressed], provided the loss doesn't end up too obvious in use.
A 'bitmap' is or rather - can be loss-less, as at 32bit ALL data is retained....but the image 'weight' is huge.
Compression of an image re-reads the full bitmap and approximates/combines adjacent pixels of same [or similar - depending] colour.
Working with such an image afterwards compounds the loss [of detail].
It's a one-time-only deal...