Most of what you think you are seeing at any moment is only imagined.
The retina is not uniform. Most photosensitive cells (cones) are near the centre of vision. Peripheral vision has little resolving power. Can't make out fine details. The reality of this is much more extreme than it subjectively feels like. The eye doesn't actually have pixels but if it did they'd all be focused at the centre. Like an image where 10% of the area in the middle had 80% of the pixels.
At the centre of vision the eye has enough resolving power to make out the tiny star shapes and see that they are rotating. Outside of that narrow zone in the peripheral vision they're perceived as coloured blobs, at best. Normally your brain would make this transparent to you. But this is an unusual pattern. Your visual cortex doesn't realize all the stars should be rotating. So only the ones you can actually see at any one instant seem to rotate.
Try to look at an object in the room with you, such as a lamp, without looking at it directly. Observe it out of the corner of your eye. The more you try, the less sharply defined it will seem. At the very edge of your vision you're only getting a handful of pixels worth of colour information. But because you know it is a lamp, it has the sharpness of a lamp's definition even though you cannot actually see that definition without directly looking at it. That's a related illusion.
This is why the eye scans constantly in those micro-jerking motions known as saccades. If a face were to pop up on your display, it would feel like a single instant recognition of a person. But before you experience that the eye would scan over the eyes, mouth, nose and so on, several times, in sharp flicking motions, over about 100 milliseconds, and these dozen or so little snapshots, as it were, would be stitched together into the whole image of a face. Even though only a tiny slice of the eye, or the nose, etc. can actually be seen at any one time, you perceive the whole face.
This illusion hacks that and reveals how narrow our high-resolution vision really is. The whole visual field feels rather high resolution. But only that tiny spot where they rotate actually is.
> Most of what you think you are seeing at any moment is only imagined
For some great illustrations of this I recommend people take a look at the Nova episode "Your Brain: Perception Deception" [1]. Nova episodes usually are only available to watch for free for a couple months after they air, then you have to be a PBS contributor, but occasionally old episodes become temporarily available. This one happens to be available now, with the video embedded at the link I gave.
The whole thing is worth watching, but for the material most directly relevant to this you can start at the 7:13 minute mark, where it briefly discusses a well known optical illusion and why it works, and then looks at the question of if you brain can fake your perception so much to make that illusion work, just how much of what you see in normal scenes is real? Then they go into how you only have about 1 degree of high resolution vision at the center.
A couple minutes after that there is a demo showing an interesting way to exploit that. They put the host in an eye tracker that can figure out where she is looking. Then they have her read some text on a monitor and she has no trouble reading. But when the camera shows us the monitor we see that almost the text is just the letter 'X' with small groups of letters briefly switching to other letters and then going back to 'X'.
When we read our eyes don't smoothly scan the text. They actually look at a fixed point for a moment, then jump to another fixed point, and so on (the "saccades" mentioned in the above comment). This isn't just when reading. This is how we look at everything.
She can read the text because the eye tracker is figuring out where she is looking a the software shows the real text there. As soon as her eyes jump that text goes back to 'X'.
When we look at the text and aren't looking exactly where she is we might detect the change to normal text where she is looking but we can't read it because most of the time it will be outside of the small high resolution center of our vision. By the time we can move our eyes down to it her eyes have jumped and that text is back to 'X'.
We don't consciously control the jumps so we can't sync ours to hers, so at best all we can hope for is to occasionally get lucky and maybe get a couple words now and then.
Is it a problem if the shape that I see rising stars in is not round? I get an upside down L shaped mass of rotating stars, no matter where I look in the image.
The retina is not uniform. Most photosensitive cells (cones) are near the centre of vision. Peripheral vision has little resolving power. Can't make out fine details. The reality of this is much more extreme than it subjectively feels like. The eye doesn't actually have pixels but if it did they'd all be focused at the centre. Like an image where 10% of the area in the middle had 80% of the pixels.
At the centre of vision the eye has enough resolving power to make out the tiny star shapes and see that they are rotating. Outside of that narrow zone in the peripheral vision they're perceived as coloured blobs, at best. Normally your brain would make this transparent to you. But this is an unusual pattern. Your visual cortex doesn't realize all the stars should be rotating. So only the ones you can actually see at any one instant seem to rotate.
Try to look at an object in the room with you, such as a lamp, without looking at it directly. Observe it out of the corner of your eye. The more you try, the less sharply defined it will seem. At the very edge of your vision you're only getting a handful of pixels worth of colour information. But because you know it is a lamp, it has the sharpness of a lamp's definition even though you cannot actually see that definition without directly looking at it. That's a related illusion.
This is why the eye scans constantly in those micro-jerking motions known as saccades. If a face were to pop up on your display, it would feel like a single instant recognition of a person. But before you experience that the eye would scan over the eyes, mouth, nose and so on, several times, in sharp flicking motions, over about 100 milliseconds, and these dozen or so little snapshots, as it were, would be stitched together into the whole image of a face. Even though only a tiny slice of the eye, or the nose, etc. can actually be seen at any one time, you perceive the whole face.
This illusion hacks that and reveals how narrow our high-resolution vision really is. The whole visual field feels rather high resolution. But only that tiny spot where they rotate actually is.