Info: Comb Filter Testing — Page 5

metaleonid said:

Compro VideoMate TV Ultra

I can't find one for sale anywhere. Where did you buy yours?

metaleonid said:

I bought it in 2004. I think it is this one. http://www.comprousa.com/en/product/vmtvm800-hardware.html

But why would you want it?

I'd like to test it out, mostly.

metaleonid said:

3d comb filter is good. But for some reason overall image is blurrier. I can't explain why. I can test for you while it's still in my computer. I plan on taking it out.

Thanks, I'd appreciate that.

Did you try calibrating the "sharpness" setting (if it has one)?

It sometimes shows up in the "proc amp" settings.

It looks like DIAMOND released a new driver for the DIAMOND ATI Theater 750 PCIE (TVW750PCIE) on 2013-09-10.

Get it at the following hyperlink under "Support & Downloads": http://www.diamondmm.com/tvw750pcie-diamond-pcie-hd-tv-tuner-card.html

Please tell us if it fixes the AGC issue!

metaleonid said:

Also AFAIK the AGC problem is in hardware, not in driver.

Possibly. It could also be the driver dynamically modifying a register value on the chip in response to its own analysis.

drngr said:

I'm always angry when I watch my parents' little 13" HDTV because of how good the SD video quality is. The 3D comb filter is just ridiculously effective, not only with LD but also with super-noisy analog cable signals (and the noise filter turned off). The only rainbows I saw last time I was there were on a black-and-white striped shirt, and they went away as soon as the character stopped moving.

They got this thing on sale for under $75, meanwhile there are $500 capture cards still made with 2D-only filters.

It might be possible to rip the board out and use it standalone if we can figure out how to convert LVDS.

I'm currently working on writing my own software 3D motion-adaptive comb filter. The prototype is in Python; the release code will be in C (ANSI C89/ISO C90) for Vapoursynth.

My #1 goal is to comb as much as possible without introducing artifacts.

The results, so far, are extremely promising.

There is some hope for a virtualdub filter, so it could work during video capture?

_,,,^..^,,,_ said:

There is some hope for a virtualdub filter, so it could work during video capture?

It would probably be too slow. The optical flow algorithm I'm using is pretty exhaustive and I call it several times.

I'm shooting for quality and low artifact risk rather than speed.

OK, got it... please post your work advancements here!

HF9G vs R7G using the ati theater750 PCI-e card. The raw video files are available on a.b.sw. As always, view the full sized picture. Left click to view full sized.
Test Pattern Frame #1

Test Pattern Frame #2

Test Pattern Frame #3

Test Pattern Frame #4

Video Sweep Frame #1

Video Sweep Frame #2

Video Sweep Frame #3

Video Sweep Frame #4

Color Bars Frame #1

Color Bars Frame #2

Magenta Frame #1

Blue Frame #1

Green Frame #1

I've been studying the avisynth branch of the Bifrost filter. Even though it doesn't use a sophisticated motion estimation algorithm (e.g., optical flow) I've found that it does a better job than pretty much every other software 3DCF* out there currently, TComb included.

I've been torture-testing it too. At reasonable settings it's hard to get it to artifact. It has an altclip parameter which gives you the ability to provide a different clip as an argument for the blocks it classifies as being "in motion". You can use this to process only the motion blocks with a different filter; e.g., the low-pass filter I've mentioned in other threads.

I've been using:

    c.Bifrost(altclip=c.Spline16Resize(180, 480).Spline16Resize(720, 480))

This allows me to 3DCF static blocks and LPF motion blocks only. It works really well for the footage I'm working with at the moment.

You can considerably improve its motion detection by getting rid of all dot crawl first since it uses the luma plane to determine if a block is in motion or not and dot crawl screws that up.

I suggest capturing a S&W zone plate with your setup in order to find the best values to provide as arguments for the luma_thresh and variation parameters.

(* Keep in mind that this is for chroma only. You'll have to go elsewhere for dot crawl.)

Thanks for the information.  I will give that a try. I have already installed bifrost awhile ago but haven't played with it. I do something similar with tcomb but since it doesn't have an alt clip parameter I do this:

s1d1=avisource(1d comb output from ld-decode)

s1d2=avisource(2d comb output from ld-decode)

s1d3=s1d1.tcomb(2,4,5,5,6,false,12)

s1d3mask=s1d1.tcomb(2,4,5,5,6,true,12).greyscale().invert()

s1adapt=overlay(s1d3,s1d2,mask=s1d3mask,mode="blend",opacity=1)

It's kind of a hack but works like it should.

So with bifrost it might be best to use a clip which has already been 2d combed? If so, a person could use tcomb/a good 2d comb to eliminate dot crawl and then bifrost to eliminate rainbowing.