| Author |
Message |
barcodude
Joined: 14 Jun 2014
Posts: 169
Location: The Netherlands
|
| Posted: Sat Jan 02, 2016 2:59 pm Post subject: |
|
|
maybe a stupid question but wasn't this already done by faroudja with true life enhancement in pixelmagic systems like the crystalio an the pe-1000 pro 
_________________
Epson EMP TW700 - Projecta HomeScreen Deluxe- Denon a/v receiver - Mission highrisers -klipsch subwoofer - Samsung Bluray player
|
|
| Back to top |
|
 |
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Sat Jan 02, 2016 3:00 pm Post subject: |
|
|
I do not know that technology but might very well be true. I am finally going for a darbee solution to try like the oppo 103d.
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
xmob135lc
Joined: 15 Sep 2012
Posts: 80
|
| Posted: Sun Jan 03, 2016 4:06 am Post subject: |
|
|
| barcodude wrote: | | maybe a stupid question but wasn't this already done by faroudja with true life enhancement in pixelmagic systems like the crystalio an the pe-1000 pro |
| Quote: |
TrueLife® Enhancement identifies patterns of transition which contain important detail in an image.
|
No, cause that's the
high frequency contrast "accentuation" - sort of like boosting audible(?) 20khz when your tweeter falls rapidly above 15khz, making matters worse , and also it's as quantized/digital as it can get because they select frequencies instead of smooth filtering- that's only great for test signals. I'm also aware of these they used to (fail to ) combat the temporal blur of slow LCD-s the same way, (instead of using better lcd mixture ) there it's called "overdrive" , very dubious and gimmicky.
So again, the higher the peaks are increased , the "valleys" also come up in level because of the gaussian "skirt" like luminance distribution, called "flare" , and that contributes to blurring, it's the same wrong, just wrong in a different way ,
probably even when "everything is measured", that sounds like the inverse-FFt "room correction" in audio -dubious claims and usually only correct in a spot with the correct test signal.
My suggestion is different, it's actually like having extra pixel shape definition, because there ARE new "negative" / "blank" values (actually won't be because ideally it cancels out so OVERALL, there is NO alteration of pixel intensities here), not an accentuation, but making the space for the "skirts" to take up.
Add to the list: unequal horizontal/vertical MTF , depending on tube face location,so the usual suspect "textbook" Lanczos filter probably isn't even enough, has to be VERY non-uniform.
Worth a try, if there is bandwidth to spare, the oddball resolution at 60hz like 2203x1124 might give at least equal horizontal/vertical MTF.
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Sun Jan 03, 2016 5:29 pm Post subject: |
|
|
I tried the complex sharpen 2 shader in MPC-HC
http://forum.doom9.org/showthread.php?t=150798
It is an unsharp mask so not an edge enhancement so worth a try.
I must say on movie material it looks very sharp and nice. I found it hard to go back to the no shader option. With a testpattern it clearly improved on the white lines but little on the 1:1. I did also see some ringing on the numbers in the testpatterns but in the movie material it simple looked good. So try it for fun if you have a pc.
nb unsharp mask does not work with FFT or with discrete frequencies. It uses a gaussian blur or low pass filter to enhance the high frequencies.
https://en.wikipedia.org/wiki/Gaussian_blur
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
xmob135lc
Joined: 15 Sep 2012
Posts: 80
|
| Posted: Sun Jan 03, 2016 8:19 pm Post subject: |
|
|
I think this audio analogy works,
if you have a loudspeaker with a tweeter that has practically no good response beyond 15khz,usually it's not a solution to use whatever 256 dB/oct FFt filter and crank it up , because that's artificial/ stressful and what not...
But they used to add another tweeter,&/or make the soundwaves directional with a horn.
What I said is a bit kind of both , new pixels /tweeter,/ and outskirt is shaped /directed/ out. Even after this, accentuation/cranking can be stressful, because you end up where you started (skirts).
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Mon Jan 04, 2016 2:38 pm Post subject: |
|
|
So you are saying that with CRT there is a kind of interpolation between pixels. So if one pixel is white the next black than in between it is grey. Now with the lanczos oversampling there are zero vallues added in between the pixels. I am thinking that if the beam has to go from black to white than an in between black improves the black if the bandwidth is not the limitations. I mean if the beam gets an even shorter time to rise that could be a problem. If that is not a problem than the overlap of the gaussian's is less. Resulting in better MTF.
So you say with the new neckboards bandwidth should not be a limitation. I have a pretty fast videochain so it might be worth to try a lanczos oversampling to a higher resolution.
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
xmob135lc
Joined: 15 Sep 2012
Posts: 80
|
| Posted: Mon Jan 04, 2016 9:58 pm Post subject: |
|
|
| Quote: |
I mean if the beam gets an even shorter time to rise that could be a problem. If that is not a problem than the overlap of the gaussian's is less. |
it's by definition too large with gaussian, and the higher the peak luminance the larger the skirt is, so trying to solve this with contrast accentuation complicates the situation.
Let's just refer to the CRT's blur as overlap in effect it's "gaussian blur".
| Quote: | The Gaussian filter is a very special filter that generates that same 'bell curve' shape in the frequency domain. This makes it very useful as an image filter as it guarantees a good removal of this high frequency noise in a highly controllable way.
However if you examine the filter graph, you will see that at a distance of one pixel from the sampling point, you have a non-zero value. In fact it is quite a high value indeed. This results in a huge amount of blurring of the individual pixels, even when no resize is actually performed |
in this case it's not removing high frequency noise , but information.
Now see the Lanczos
Again non-zero value, yet this time it's a negative!
So, if you add these together, and the two match ,you get a "NULL", hence an useful "blanking" property...
Obviously you have to make sure the two match and this requires you to know the negative value necessary across the tube face to get the true null, so for the best result you need a non-uniform Lanczos filter, not a 'Sinc' for example since that gives long ringing and can push the sides of the gaussian skirt "up" .
| Quote: |
(...) a 2-lobed 'Lanczos2' filter (Lanczos with a default lobes of 2, added for easy user selection) has also been found to be popular, as it will avoid the positive Ringing Artefacts that can be generated by Windowed-Sinc filters.
|
If the two match , so there is nulling in practice , you get the "spline" type pixel shape.
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Mon Jan 04, 2016 10:45 pm Post subject: |
|
|
Ok in my own language. You have the samples that are ideal spikes they get convolved with gaussian blur from the beam but you want to mathematically convolve with lanczos shapes to make the result more like a spline. Getting the correct parameters is the problem but even if it does not blank perfect it should still improve the mtf.
Sounds smart.
I am still thinking that since the beam shape is a gaussian blur that an unsharp mask that is an inverse gaussian blur is the simplest way. Ok the blur will get bigger with sharpening but the net effect is still sharper.
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
xmob135lc
Joined: 15 Sep 2012
Posts: 80
|
| Posted: Tue Jan 05, 2016 1:27 am Post subject: |
|
|
| Quote: | | gaussian blur from the beam |
No, probably the direct contributor is the phosphor powder, there's a "scattering" effect taking place.
If it was for the beam there would be a tight spot.
Seems like you just want to throw off my hard earned "null" . 
The thing you can do , is to budget for a degree of contrast accentuation by choosing a Lanczos type that goes deeper at 1.0 than usual,
but there will be a price: ringing.
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Tue Jan 05, 2016 10:28 am Post subject: |
|
|
| xmob135lc wrote: | | Quote: | | gaussian blur from the beam |
No, probably the direct contributor is the phosphor powder, there's a "scattering" effect taking place.
If it was for the beam there would be a tight spot.
Seems like you just want to throw off my hard earned "null" .
The thing you can do , is to budget for a degree of contrast accentuation by choosing a Lanczos type that goes deeper at 1.0 than usual,
but there will be a price: ringing. |
I once worked with electron microscopes and we could improve the image by catching the secondary scattered electrons with a grid around the camera. Also the field emission guns where much sharper than the thermal emission guns perhaps to sharp for CRT tubes?
Anyway we are now getting to hardware improvements and than we should discuss the modified video chain too.
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Tue Jan 05, 2016 11:31 am Post subject: |
|
|
When I read this post again it seems to me that digital does not have perfect MTF either.
http://www.videovantage.com/?p=819
Even the RS35 only has 79% horizontal MTF.
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
xmob135lc
Joined: 15 Sep 2012
Posts: 80
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Thu Jan 07, 2016 9:05 pm Post subject: |
|
|
Just an idea would it in principle not be possible to use a pwm signal like the signal coming from a digital amplifier on the CRT beam? I mean why do we make it analogue? A black pixel would be no pulse and a grey pixel would be a small pulse and white would be a full pulse. So we would not modulate the amplitude of the beam like we do now but we would modulate the duration of the on off of the beam, the pulse width.
Just in principle would such a digital pwm modulated CRT be possible? If so it would not suffer from video bandwidth and it would be possible to put blanking between the pixels.
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
xmob135lc
Joined: 15 Sep 2012
Posts: 80
|
| Posted: Thu Jan 07, 2016 10:21 pm Post subject: |
|
|
Not sure if you need actual hardware blanking on CRTs , is even 1:1 pixel mapping so important ?
If you think about it , the higher the incoming resolution the less suspectible the visible scaling artefacts , and fhd is already plenty high.
They had the sanity to show 5pixels alternating in the MTF test too, 1-1 pixel alternation is not real-world like signal, also computer graphics are anti aliased for the better part.
Better would be the latest scatter-free LED phosphors .
| Quote: | | Scatter-free phosphors promise to suppress the scattering loss of conventional micro-size powder phosphors |
| Quote: | | well-known, the particle sizes of phosphors do effect the strength of the scattering |
p22 green is the worst offender, energy buildup is slow and release is also bad, no rare earth there , so the ordinary LED stuff is worlds better not to mention the new ones probably can be hit even harder without giving in early.
| Quote: | | high damage threshold, good thermal conductivity, good chemical stability |
|
|
| Back to top |
|
|
cmjohnson
Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s
|
| Posted: Wed Jan 13, 2016 8:50 pm Post subject: |
|
|
|
I think that was something that was done by the Teranex processors. Remember those?
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
|
| Back to top |
|
|
mc86
Joined: 20 Sep 2008
Posts: 767
Location: pittsburgh, pa
TV/Projector: ECP 4500 (Vidikron box), ECP4500+, wanting 07MS/07MTS, evaluating pc soft-blend
|
| Posted: Thu Jan 14, 2016 1:48 am Post subject: |
|
|
Interesting stuff. Sadly, I can only add one humorous reflection here: A few years ago I bought an image analysis add-on (for LabVIEW) that included a function called the "Wiener Deconvolution"...I was in a sleep-deprived state around 2AM in my office when I stumbled across the function. After repeating it to myself a number of times, I worked myself into laughing so hard I was in tears. There was no one with me, nothing readily funny around (just equations on my screen), etc. when in walks the cleaning lady to empty my trash! I waved her off in my uncontrollable state and, with strange looks, she eagerly scooted out. I can only imagine how insane she must have thought me.
Of course, all us guys on here know how to carry out this function.
cheers,
matt
|
|
| Back to top |
|
|
xmob135lc
Joined: 15 Sep 2012
Posts: 80
|
| Posted: Thu Jan 14, 2016 7:27 am Post subject: |
|
|
it depends on composition , if it's powder / bulk then it will scatter,
if it's ordered in the right way then it will behave like a laser spot. I think the usual coating procedure rules out anything but powder type composition, maybe not.
| Quote: | | Large micro-size cube phosphors (~100 μm) are newly designed and prepared as scatter-free phosphors, combining the two scatter-free conditions of particles based on Mie’s scattering theory; the grain size or grain boundary was smaller than 50 nm and the particle size was larger than 30 μm. |
https://en.wikipedia.org/wiki/Mie_scattering
|
|
| Back to top |
|
|
redfox001
Joined: 16 Mar 2009
Posts: 2257
Location: The Netherlands
|
| Posted: Fri Feb 19, 2016 5:58 pm Post subject: |
|
|
| xmob135lc wrote: | I think this audio analogy works,
if you have a loudspeaker with a tweeter that has practically no good response beyond 15khz,usually it's not a solution to use whatever 256 dB/oct FFt filter and crank it up , because that's artificial/ stressful and what not...
But they used to add another tweeter,&/or make the soundwaves directional with a horn.
What I said is a bit kind of both , new pixels /tweeter,/ and outskirt is shaped /directed/ out. Even after this, accentuation/cranking can be stressful, because you end up where you started (skirts). |
So upscaling hd to a higher resolution and pulling in between pixel vallues to zero. If there is enough bandwidth this would give a much sharper focus. Better mtf. When I have some bandwidth to spare I will try this trick to realise the pulsed CRT without hardware change. Vertical and horizontal can be made much less gaussian convoluted even without special phosphor 
_________________
701s->runco933->8500ultra->cinemax->9500mp->919 splitpack + cinemax
|
|
| Back to top |
|
|
cmjohnson
Joined: 03 Apr 2006
Posts: 5180
Location: Buried under G90s
|
| Posted: Fri Feb 19, 2016 11:32 pm Post subject: |
|
|
Ultimately it comes down to the characteristics of the phosphor and the beam profile. We can get a narrower beam (to some extent) with better focus magnetics.
But once a tube is made, the ultimate smallest electron beam spot size and the phosphor grain size is a fixed parameter, and you not be able to do anything to improve the MTF beyond those built-in limitations in the CRT itself.
No fancy algorithm or custom electronics is going to change that.
Last edited by cmjohnson on Sat Feb 20, 2016 3:33 pm; edited 1 time in total
|
|
| Back to top |
|
|
|
|
