Mainly, I think you practice like you would any skill, controlling motivation, repetition, difficulty, etc. I would be very surprised if sustained training yielded no improvement.

I know I've read books on it before. Never applied myself for any length of time, so can't give a fair assessment.

And for your own training, I don't think 2 was the right thing. That might help you to hold an image, but I don't see it helping you generate the image to hold. 3 doesn't look helpful either - I don't think logicking it out would help.

And even 1 is questionable, as you're trying to maintain pieces of a sleep state. That might allow you to visualize if you can pull it off generating enough of a sleep state.

Try to do colors. Then basic shapes. Try eyes open, eyes closed. I find it easier to in the basic Matrix white space. Try to get a 3d image, instead of 2d. Maybe simple isn't the right way. Maybe a natural scene with detail is easier. And also, see about adding other sensory modalities. I'm probably best at "visualizing" sound.

A lot of dance and movement theorists recommend "visualizing" body movement as well.

Basically, you're just learning how to control and train your nervous system, so I think a lot of movement theory would transfer. I wonder if Feldenkrais ever had anything to say on visual visualization. He's very good on movement and learning theories.