If it's on one layer then "removing" the red paint will leave white pixels, as there is nothing underneath to see. Images are made of pixels and when new ones are put down they replace the former ones. If you are trying to recover prior information, that cannot be achieved.
And just to clarify: In a flat pixel image (an image without Layers, like jpgs are by definition) there are no hidden pixels behind the pixels one can see.
So while if may be possible to create meaningful content to fill image space that is occupied by an object (in some cases Content-Aware Fill with a rough Selection may suffice for that) or use Adjustment Layers to reduce the effect of transparent objects on the objects behind them it is not really possible to deduce plainly invisible objects.