CurtPalme.com Home Theater Forum

[opv]

Hi,
Do you know if MTF mechanisms are applied in displays today?
Do you know of such mechanism for CRT projectors ( a VP feature, maybe)?
To my understanding, such mechanism has the potential to do wonders for CRT sharpness.

[Phoenixed]

Isn't MTF, mean time between failure?

[opv]

Funny, I wonder if you can compensate that...

I meant MTF=modulation transfer function

[Tim in Phoenix]

Hello

Are we supposed to know what that is? Explain pleeze.


.

[Spanky Ham]

I am not sure what you mean by MTF compensation. Do you have a link to anything using this?


MTF was explained years ago by tse on a thread over at AVS. The MTF of CRTs is pretty low.

[Nashou66]

WTF is MTF?

here ya go!!

http://www.avsforum.com/t/633597/wtf-is-mtf



Nashou

[opv]

Well,
MTF compensation is a technique that compensates the known MTF of a component in a chain in order to get back details that were lost due to the non optimal MTF.
It's used to improve pictures from a sensor with a known MTF.
It is also used in digital printers to improve the output by precoding the input image.

I'm trying to find out weather such a technique, is used in display devices.

[Spanky Ham]

I just did a search and this thread comes on the first page of google.

I don't see how you could increase the MTF, since the weak link is the tube and optics. Maybe I am missing something here.

[opv]

Hi Spanky,
You can improve the MTF by applying precoding on the source.
The MTF can be modeled as a linear 2 dimensional pulse shape filter.
If you measure the impulse response of this pulse shape, you can apply the inverse 2D filter on the source file.
It's similar to pre-equalization technique used in digital communication.

[Spanky Ham]

I think I see what you are getting at. Tim (videograbber) can comment on this, but I think this is what the Darbee is doing.

[opv]

Sorry, but I don't think that's what Darbee is doing due to the simple fact that the Darbee doesn't measure the MTF of the display.
I think the Darbee algorithm is some sort of advanced local contrast starching.

[xmob135lc]

The EASY way to do it ,eg. without getting intimate with hardware , is a special software rescaler , working towards a "single" native resolution,
and
resize EVERY incoming spatial resolution in a 'really special' way to one or two refresh rate , either multiple of 30 or 25 plus multiple of 24

and using maybe one of the following type of filters see here:

http://www.imagemagick.org/Usage/filter/nicolas/

http://www.imagemagick.org/Usage/filter/nicolas/enlarged_sigmoidize.png

<--- the "blackened" value on Lanczos etc. looks like could work as a "blanking" feature, preventing one pixel value to modulate the one next to it (is basically what we want).

[redfox001]

One thing in my mind perhaps we could simple do a spatial fourier analysis of one line and increase the contrast on the higher frequencies? With the simple smpte testpattern the 1:1 would be increased in contrast and the 2:2 less. Result better mtf?

[redfox001]

The MTF compensation processing involves two one-dimensional Fourier transform filters using measured vertical and horizontal MTF functions as the bases for constructing filters, she said.

http://www.physicianspractice.com/articles/crt-compensation-tool-aids-detection-performance

[xmob135lc]

I think it boils down to the peculiarities in the actual video chain, so I'd suggest a "native resolution" eg. ~samey pixel clock and refresh values close eg. somewehere around 60-72-75 , probably important to give up multi-sync / resolution for a higher degree of improvement.

Increased contrast sounds like it'd still result in overlap between pixel values hence ,not a good solution, or worse, going straight for erroneous output if we can't match actual "peculiarities ", scaling with pseudo-blanking however seems a safe-bet.

[redfox001]

Ok you might be right have to understand better how pixels overlap. For the simple contrast aproach how about unsharp mask from ffdshow filters?

http://ffdshow-tryout.sourceforge.net/wiki/video:sharpen

Explanation. It takes a blured image with less high frequency contrast and subtracts from the original. Result a image with enhanced high frequency contrast.

I am going to test that filter on the smpte testpattern soon. See if we can make a perfect 1:1 with a limited mtf crt.

[redfox001]

[xmob135lc]

Wonder how 1920x1080 (and less) upsampled to 2560x1080 with sharpening scaler, then shown as 16:9 on the CRT probably could work, especially with recent wave of neckboards, that's some easy "blanking" for sure.

I think increased contrast is asking for trouble : too easy to get erroneous output .

[xmob135lc]

http://www.imagemagick.org/Usage/filter/#artefacts

great rundown here, note how the gaussian pixel shape is non-zero at the one pixel distance, and the Lanczos type filter is suited to set it to zero, I think it's a very neat solution (the pseudo/"blanking" arrangement - it can even selectively null overlap per pixel and store correction values for the entire tube face ) , yet it took a while for suitable high-perfomance neckboards to appear .

Also 60hz in fast-retrace mode should be o.k. (eg. keep drawing until the end of frame)

[redfox001]

I think we would need both. The scaling and blanking might be espesscially usefull for Marquee. I have seen a lot of sharpness appearing when a Barco 909 is focussed and flare is reduced with the astig system. That must be the kind of mtf you are talking about and certainly is a crt weakness. The contrast enhancement is to compensate for the limited bandwidth in the video circuit. Limited rise time of the intensity.

https://en.wikipedia.org/wiki/Unsharp_masking

Also deconvolution could be applied.