- This topic has 12 replies, 7 voices, and was last updated Nov 2, 2021 4:43pm by
.
-
Posts
-
I always wondered (exspecially now after the new vorpx sharpening has been introduced) how small a pixel can be before it doesnt make sense to raise a headsets hardware resolution any futher. I came to this because i wondered how it would look like when shrinking down a 19″ monitor to 2″. I mean for example 1080p on 19″ looks quite sharp, shrinking that down to 2″ must even look more sharper and it seems it does (or does it not beacuse of FOV ? I only slowly get an idea by pushing that cinema screen back to a 19″ size). So 2880p shrunk down to 1 oder 2″ must even look more sharp and so on. Now since a Vive Pro for example has a hardware resolution of 1440×1600 per eye,rasing resolution way above that makes the image sharper and sharper as more i raise my resolution setting. 2880p for example. Now there are headsets with 2440×2440 for example. I wonder if it would make sense to buy such a Headset while i still can keep gaining resolution to the limit of my PC with one with less hardware resolution.
What i basically want to find out whats the limit of the human eye. 4k, 8, 50 k and why can image quality visually be raised by going over the hardware resolution of a headset which i find is weird ? Any thoughts in theory on this matter ?
Average human eye resolution is roughly around 60 PPD (pixels per degree), little more with good vision. Current consumer headsets are 15-20 PPD. You can calculate it yourself by taking the headsets resolution and dividing it by the FOV (e.g. 1440px / 100° -> 14.4 PPD). High-end headsets like the Varjo Aero or the new 12K Pixmax go up to around 30 PPD. Varjo XR3/VR3 already reach 70 PPD using a special display setup (tiny high resolution screen in the center, with the surrounding being a lower resolution screen).
The PPD of a monitor, TV or cinema is roughly around 30-100 PPD, depending on viewing distance and if it’s a HD or 4K display (FOV is around 30-60°).
why can image quality visually be raised by going over the hardware resolution
That’s called “supersampling”, instead of having your computer render one pixel per display pixel, you let it render multiple and average them together before sending them to the display. This ends up looking better as it reduces “aliasing”, i.e. the little pixel stair-steps you often see in computer games graphics without anti-aliasing enabled. If you enable anti-aliasing/supersampling, the stairsteps will be smoothed out.
supersampling does more than just removing aliasing, sharpness comes from details and redering software/hardware is designed to be efficent, to not render more than needed so when you render natively you just dont get details past render resolution. What if there is a needle on a table but its sub pixel size if you render natively, so it might not be visible at all, wile at higher resolution even resized back it would remain as a half transaprent line of pixels indicating its smaller than pixel size.
Furthermore some engines dont render small details past certain distance if it is deemed to not be visible by end user and sometimes its wrong. I had bullet holes disappear on walls much sooner in vr games when playing 100% supersampling and then magically appear if I raise it up.
TL/DR: rendering higher resolution shows more detail and resizin it back to output still preserves all that detail = thats where sharpness comes from, better antialiasing is just a side effect.As for headsets – a review of that Varjo Aero (2880 x 2720 px per eye) pixels cant be seen, so I assume limit will be vorpx screen resolution after that and screen resolution pixel spotting can varry, in full vr for example pixels will be noticeable even if you render at those 2880 x 2720
Average human eye resolution is roughly around 60 PPD (pixels per degree),
intersting to read, thanks !
That’s called “supersampling”,
I assume the better image when rendering over the headset resolution is a result of sort of a “wiping out AA more and even more”, not of providing me with more pixels. Having this in mind i wonder how far i could go in resolution in theory on a 1440×1600 display to maximize visuals. I usually play my games around 3200×2880, today i did a quick test on Avatar and Fallout 3 and there was still a noticable improvement in sharpnes and claryity on small detailed objects like letters on guns or stickers on uniforms for example when gaining resolution to 4400×3240 (my current monitor cant take more ;-). The image looks so clear and clean in that resolution so i wonder if a 5 or 8K display ever would provide me with better visuals at the same (locally maximized) resolution.
I think the effect of supersampling is explained differently. When your head is completely still, you will always see the screen door effect, no matter how high the supersampling is. But when you move your head even slightly, your brain processes the incoming information and reconstructs a high-resolution image. So, there is actually no limit to supersampling resolution (but does not make sense go beyond human eye resolution). It is roughly like looking through an actual screen door: if you look through it at a low-quality picture, you will always see a low-quality picture and grains, but when you look at a high-resolution picture and move your head, the screen door kind of disappears.
Very intersting. Funny sidenote: I experimented with 3D a lot in the past X years. One day fooling around with an empty toilet paper roll i noticed that the brain is able to constuct a 3D image with just one eye. The trick was to blend out everything else exept the 3D object that was displayed on my monitor for the other eye. I named it “Monofokularstereoskopie”. (german)
According to Vsauce, 7 Megapixels for the center of focus plus 1 megapixel for the rest, while it is a simplification the data provided is solid, with resolutions of ~300 ppi (pixels per inch) making each pixel unresolved for normal vision.
I experimented with 3D a lot in the past X years. One day fooling around with an empty toilet paper roll i noticed that the brain is able to constuct a 3D image with just one eye.
Similar thing I noticed with VR360-2D videos, watching them with both eyes feels all wrong, people are huge and everything looks off. Watching the same video with only one eye makes it look almost 3D, people have the right size and everything feels far more believable than with both eyes. Seems like removing the incorrect depth information from the second eye makes the brain switch over to another depth guessing algorithm that works much better.
All that makes me wonder if there would be any further need of increase in resoultion once we made it close or just over the Screendoor Barrier (BTW: great Video !) The brain factor seems to be more important as we may think.
one thing to consider, is that hmd panel resolution doesnt match 1:1 with whats actually rendered on it, some distortion have to be baked in to offset the lens.
Brain factor is a damn thing that sometimes convinces you there is kinda depth to a game when there is none.
Thats one reason I wish there was depth buffer mode to check if there is any data at all.
There is a toggle to switch 3d on and off, but even with it brain can play tricks.
While latest HMDs almost or fully eliminate the SDE through high resolution, the SDE itself is not caused by low resolution, but by gaps between pixels.
Another approach of eliminating SDE would be for pixels to fill a bigger area. There were various attempts with using different sub-pixel arrangements, micro lenses, etc.
Even without these the gaps are significantly smaller than pixels. They’re also washed out by light scattering in the lenses. So in order to eliminate the SDE you don’t need to match the eye’s “resolution” in terms of pixels, you only need to make the gaps disappear.
I would bet that even with something like Varjo Aero, Pimax 8K etc., even if there’s functionally no SDE, you’re still not able to resolve some details that you would be able to in real life, for example on far away objects.
i do have a pimax 4k and 8k and while the 8k feels very clear vs the 4k, even with 8k its not yet at the point where small details are sharp, actual vr games feels stupidly sharp (and the wtf fov level of the pimax 8k often mean that you can see stuff loading to the edges) but games not made for vr usually have smaller text/ui and on those you can see the pixel/gaps.
Still more pixel density should allow for improvements, I’m curious to see how its gonna look like with the 12k resolution panel that pimax announced for their next headset, altho at the same time, i doubt my 2080 could power it considering even the 8k is pushing it in some scenarios heh.
- You must be logged in to reply to this topic.