What exactly are "pixels of motion blur"? [GtG blur versus MPRT blur]

Advanced display talk, display hackers, advanced game programmers, scientists, display researchers, display manufacturers, vision researchers & Advanced Display Articles on Blur Busters. The masters on Blur Busters.
Post Reply
millysoose
Posts: 2
Joined: 25 Oct 2021, 01:01

What exactly are "pixels of motion blur"? [GtG blur versus MPRT blur]

Post by millysoose » 25 Oct 2021, 02:46

I'm trying to wrap my head around a statement like this (taken from the excellent article on Blur Buster's Law): 2000 pixels/second at 16.7ms persistence = 33 pixels of motion blurring

I understand everything up to 33 pixels of motion blurring. Does that mean there will be will be a separation of 33 pixels from where the object being tracked is positioned and where your field of view is centered at the end of the refresh? In other words, between the start and end of a refresh, the (stationary) tracked object is smeared across 33 pixels in its direction of movement? Do I understand this correctly?

Where I begin to doubt my understanding is when I try to understand the "Vicious Cycle". Because, following Blur Buster's Law, the same physical motion speed produces greater pixel blurring at greater resolutions. But I think back to the eye-tracking UFO test, and how the black lines thicken as your eyes chase the UFO. Now, wouldn't those lines be just as thick on a higher resolution monitor, all else equal? Yet the higher resolution monitor has more "pixels of motion blur". But if motion blur boils down to a smeared eye-tracked object, why is it measured in pixels when pixels come in different sizes?

thatoneguy
Posts: 181
Joined: 06 Aug 2015, 17:16

Re: What exactly are "pixels of motion blur"?

Post by thatoneguy » 25 Oct 2021, 04:39

Basically divide horizontal resolution with framerate.
Say for example for 3840x2160@60fps sample-and-hold it would be 3840/60 = 64 pixels of motion blur.
With blur reduction technology you can say simulate 1ms mprt(which is equivalent to 1000fps basically) and it would instead be 3840/1000 = 3.84 pixels of motion blur.
The higher the resolution the more motion blur you have at the same framerate because you have more pixels.

The reason only horizontal resolution is used in this formula is probably because all displays scan from left to right.

Pixel sizing or shape doesn't matter much since you're not likely to see pixels from a normal viewing distance.

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: What exactly are "pixels of motion blur"? [GtG blur versus MPRT blur]

Post by Chief Blur Buster » 25 Oct 2021, 10:28

Yes, pixels of motion blur is the number of pixels between the start of the blur and the end of the blur.

A great way to self-educate about motion blur is the stutter-to-blur continuum. Motion blur is always the frame step on a sample and hold display. So if an object moves 33 pixels between frames, it will have 33 pixels of motion blur. You raise the refresh rate and frame rate to decrease sample-and-hold motion blur.

www.testufo.com/eyetracking#speed=-1



In this animation, you will see the lines are always there by looking at the stationary UFO.
But looking at the moving UFO the effect depends on the speed (frame rate).
Slow speed will always be a low frame rate in this specific test, and fast speed will always be a high frame rate in this specific test. The lines disappear when the vibration exceed your personal flicker fusion threshold (the threshold where vibrations give way to blurring). This can vary from human to human and the thresholds vary a lot due to environmental factors (and can be distorted by a slower or faster GtG pixel response). It is a very odd optical illusion if you've never seen this animation before. It is one of the most educational sample-and-hold TestUFO animations I've ever created, and perfectly demonstrates that stutters & persistence blur is exactly the same thing!

The stutter-to-blur continuum can be compared to vibrating strings -- they can sometimes vibrate too fast to see.
The motion is always stepping the distance between two lines.
If the motion is a low frame rate, it's going to stutter (vibrate), like a slow vibrating string.
If the motion is a high frame rate, it's going to blur (blend), like a fast vibrating string.

Now you're getting it -- this is blur caused by "sample and hold". Sample = a stationary frame. Hold = the frame is being displayed for a whole refresh cycle (even though your eyes are moving in an analog straight line). The held frame is blurred across your retinas as your eyes continue its analog momentum of smooth eye tracking. Display motion blur caused by persistence -- motion blur from a finite frame rate on a sample and hold display.

For simplicity, Blur Busters Law only pertains to MPRT blur, which is the distance between the two arrows (the frame step = the pixels of motion blur). If GtG=0 there is no additional afterimages, and you only see MPRT blur (one frame step worth of motion blur).

RTINGS video explains the sample-and-hold blurring caused by eye tracking very well:

phpBB [video]


We are excluding GtG-based blurring (which is additive to persistence/MPRT motion blurring from sample and hold which is very dead simple Blur Busters Law mathematics). On modern 60Hz-144Hz monitors, GtG is already mostly done within a tiny fraction of a refresh cycle (a little bit of 10% hazy ghosting results from GtG only being 90% complete in a refresh cycle). So you will essentially have a MPRT blur followed by a GtG blur (superimposed), as follows:

This does not always happen with all screens with all motions (GtG+MPRT is usually fully combined), but sometimes they clearly separate when GtG is slower than a refresh cycle.
- Blur between two rightmost arrows is dominated by MPRT
- Blur between two leftmost arrows is dominated by GtG
Image

Usually it is one continuous blur smearing over multiple refresh cycles (combined GtG+MPRT blur) but sometimes they diverge sufficiently to produce stepped blur at refresh cycle intervals, that are more dominated by MPRT (for first one) and more dominated by GtG (for the secondaries).

There are multiple artifacts to unpack here:
- Simple persistence blur (sample and hold, MPRT, persistence). Like a linear photoshop blur. The blur nearest the moving object (rightmost two arrows) will always be dominated by MPRT behavior.
- Pixel response ghosting. Complex smeary-fadey blur. This is the effect that lingers beyond a refresh cycle.
They are occuring simultaneously so it's hard to separate, but this photo is a perfect example of how sometimes you can separate the primary MPRT blur away from secondary GtG ghosts. Now it's not 100% separable as GtG blur is inside the persistence (MPRT100%) blur too. And the MPRT blur can go beyond to second refresh cycle too. Too complex to explain here in a short post -- but the pixels are essentially plotting a blur across your retinas as you track your eyes, with the simultaneous sample-and-hold effect being throttled/slightly tapedelayed by a finite pixel response time (GtG). But, here, for simplicity, most of the MPRT you see is the primary blur (first blur) and most of the GtG artifacts you see are the secondary blur (second or third). The arrows point at the frame steps so you're seeing multiple refresh cycles here, since GtG can ghost over multiple refresh cycles.

The step between two arrows is one pair of refresh cycles (pair of frames) so you're seeing 3 frames here but it could easily be 4 or 5 or 6, especially on older LCDs, or ultra-high refresh rates such as 360Hz with certain colors, or dark colors on VA panels.

You can GtG=0, but the primary MPRT blur will always be there, which is Why Does Some OLEDs Have Motion Blur?. The primary sample and hold (MPRT 100%) blur is the blur between the two rightmost arrows in the image above.

The only way to shorten MPRT is shorten frame visibility time, via:
- Flashing the frame briefer (strobing), or
- Add more consecutive frames (higher frame rate at higher frame rate).

The latter is better because it avoids the stroboscopic effect. If you strobe 1000 pixels/sec at 60Hz, you will still have the 16.7 pixels of frame stepping (stroboscopic effect) even if there's no blur, as seen at The Stroboscopic Effect of Finite Frame Rates. Now if you go 1000Hz 1000fps at the same 1000 pixels/sec motion, the frame step is only 1 pixel, and you only have 1 pixel of motion blur, and you won't be able to see stroboscopics (phantom array effects). Even the mouse cursor moving at 1000 pixels/sec will be a continuous blur instead of duplicate arrows.

Image

1000Hz isn't even the final frontier. Imagine a mouse cursor at 8000 pixels/sec on an 8K display. Even if you fix blur, you can still tell whether a display has a finite refresh rate at 8000fps 8000Hz!!! So see, the bar is very high due to Vicious Cycle Effect. Imagine! We will probably have to do many tricks (e.g. an ultra tiny 1/8000sec of GPU motion blurring below human blur visibility threshold, blurring the frame step intentionally via GPU), to fix the stroboscopic effect even when we're at retina refresh rate threshold from the perspective of motion blur.

Stroboscopic stepping effects (phantom arrays) also shows up in games too:

Image

Strobing does not fix this. It's caused by finite frame rate regardless of sample-and-hold or strobing. So even if you fix blur, you still have stroboscopics. And some people ARE really bothered by this. So instead, we like to hit two birds with one stone with blurless sample and hold -- brute force out way out of blurring and stroboscopics simultaneously.

Needless to say, we are big fans of making strobing obsolete someday, with ergonomic flicker free 1000fps+ 1000Hz+ GtG=0. The Holy Grail, as enspoused by many parts of Blur Busters research at www.blurbusters.com/area51

Related:

- Try Browser Zooming with TestUFO If you zoom (double size pixels), then 33 double-size pixels is still 33 double-size pixels of motion blur. So all of this will scale up and down. Even though it's 66 native-pixels of motion blur, same as 33 double-size-pixels of motion blur for a zoomed object to twice its original size. View www.testufo.com while zooming browser to 200% and you will see motion blur scales up proportionally, since the pixels per frame is remaining constant, and the frame step is unchanged, and motion blur is always the step between frames.

- Try Comparing 1080p and 4K with TestUFO (keep Windows DPI 100%) If you increase resolution, the motion will move slower at same pixel step. Unless you scale the motion speed up. For example, 33 pixels per frame at 1080p will be half speed at 2160p (4K) unless you double speed (66 pixels per frame). The motion blur will still be the same physical length (millimeters of motion blur) despite being twice as many pixels and being same speed. But it looks worse because a bigger difference between stationary resolution and motion resolution (at same MPRT).

The divergence between stationary resolution and motion resolution becomes worse at higher resolutions. This is a great demonstration of the Vicious Cycle Effect, where higher resolutions amplify motion blur visibility, because of bigger difference between even-sharper stationary image, versus a still-motion-blurry image. The retina refresh rate of 4K is almost an order of magnitude higher than the retina refresh rate of VHS (60fps 60Hz on an LCD). We've been stuck at 60Hz doldrums for so long.

______

Is this important?

- For some people, yes.
- Motion blur is sometimes nice to look at but it's still motion blur (forced by the display) above-and-beyond real life.
- Some people get motion blur nausea, so they need either low frame rates (stutter) or ultra high frame rates (strobing / UltraHFR).
In fact, motion blur became the dominant nausea factor for giant-FOV screens in virtual reality, so they have to strobe to reduce blur. But screens are getting bigger and bigger too, so motion blur has become somewhat more bothersome at current extreme resolutions.
- Some people get stutter nausea, so they need motion blur (higher frame rates)
- Some people are bothered by ANYTHING that looks different from real life, so the only choice is strobing (if not flicker sensitive), or UltraHFR
- Some people are bothered by flicker. This can be a big problem if you're also bothered by blur too, as you might not be able to use strobing. Ouch.
- Some people are bothered by phantom arrays (from mere distraction, all the way through nausea effects) from Stroboscopic Effect of Finite Frame Rates
- Some eyestrain from PWM dimming is from phantom arrays instead of flicker, which explains why some people get headaches from PWM dimming but never get headaches from framerate=Hz strobing.
- Ergonomic ZeroFlicker is good for most people, but it has mandatory motion blur at current contemporary refresh rates (since blur-free sample and hold via Ultra HFR is still very much only an emerging technology).
- No two people see/feel/perceive motion exactly alike. There are degrees of different motion responses, like different vision behaviors (color blindness 12% of population, mostly minor, to things like focus/prescription/etc). It's amazing how different people see motion differently, and scream about tearing, while others do not. Or not picky about stutters, while others are. Or getting motion sick but don't know why. Drilling this down shows amazing insights.

That is why ultra high frame rates at ultra high refresh rates is the holy grail (Ultra HFR = Ultra high frame rates on ultra high refresh rates = for games or for video) -- it is simultaneously flicker free, blur free, stroboscopics free. This is the path to getting closer to more universal display motion comfort for a bigger percentage of human population.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: What exactly are "pixels of motion blur"? [GtG blur versus MPRT blur]

Post by Chief Blur Buster » 26 Oct 2021, 15:30

millysoose wrote:
25 Oct 2021, 02:46
I'm trying to wrap my head around a statement like this (taken from the excellent article on Blur Buster's Law): 2000 pixels/second at 16.7ms persistence = 33 pixels of motion blurring
Hello!
Just want to check if you understood my explanation.

More pixels at higher resolutions means motion speed is slower for same pixels per second.
But if we want to keep same physical motion speed (millimeters per second, miles per hour, or whatever), you have to increase number of pixels per second to get same physical motion speed.

The Eye Tracking Animation viewed at 100% zoom and 200% zoom explains this more perfectly. Now if 200% zoom was actually sharper graphics instead of zoomed graphics...

But since the number of millimeters of motion blur is unchanged (same physical motion blur for same inches of motion blur for same inches of screen size) independent of resolution. It's an identical distance of motion blur (Equally blurry). 1080p and 4K equally blurry at same physical motion speed (double the pixels per second at 4K than 1080p). But when you pause everything, 4K is obviously sharper than 1080p.

This is another angle of explaining the Vicious Cycle Effect. Makes better sense?

I suppose I need to design a dedicated viciouscycle TestUFO animation that demonstrates a low-rez and high-rez UFO stationary (clearly differentiated), versus moving low-rez and high-rez UFOs moving at same speed (equally motion blurred and looks identical).

Basically, I'll need to create a new TestUFO pattern:
stationary low-rez and high-rez UFOs = looks different in resolution
moving low-rez and high-rez UFOs (while eye tracking) = looks identical in resolution along motion vector
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

Post Reply