Any suggestions on improving this video?

[Des]

I started this project about a month ago and have made some changes to my capturing set up and am now starting a-new:

I managed to get an unopened VHS copy of a video that I am sure was originally release as a budget line tape. It has never been issued on DVD, and probably never will.
I have captured it to avi Lagarith compression. The VCR used is a JVC SR-V101US with TBC, Video calibration and R3 all turned on. Output is s-video to a AVT-8710 then s-video from that out to the capture card which is a ATI Theater 650 Pro. The capture is done via Virtualdub.
Here is 10sec. from the raw capture avi:
http://www.zumodrive.com/share/7cQMOWQ3ZT

What I see is a lot of noise, bad contrast and a green/cyan color shift, which is all pretty consistent throughout the video when viewed on TV or via capture.

Here is my first attempt to clean it up using Virtualdub filters: IVTC - remove frame rate, autodetect and adaptive all checked; MSU De-noiser High; and finally color equalizer to improve the green/cyan shift:
http://www.zumodrive.com/share/7cQNMzhhOG

The greatest improvement seems to come from the IVTC adjustment, which I guess is to be expected. The denoise and the color correction seem to help some, but I am thinking there is still room for improvement.

I've done lots of capturing, but this is the first time I have come across a VHS I care enough about to try to improve, so I figure this is a great time to cut my teeth.

I have no experience using avisynth, but I am certainly willing to learn, especially if it has greater capabilities on something like this than Vdub does.

Thanks.

--dES

PS. If it simplifies things I can post a frame capture or two in .png instead of the avi clips, just let me know.

[Des]

<Bump>
I really am trying to learn from this tape. Any advice would be greatly appreciated.

I think a big part of it is that I don't know specifically what to look for in video to correct, or perhaps what the terminology is. My background is photographic, so I have ideas, but I'm sure they don't completely translate to video.

I also understand that 'bad is bad' and one can only do so much.

Thanks in advance.

--dES

[DB83]

This will probably not help you.

I have found that 'improvement' is in the eye of the beholder. That is to say, 'One man's improvement is another's ruination'

I do not use either vdub or avisynth. But just to illustrate my previous point I put your clip through procoder. Did not touch the frame-rate just reduced contrast by approx 10% and applied a default temporal noise reduction.

This was the result - single frame grab:

(like I said, this 'improvement' may not be to your taste.

[Des]

Hi DB83,
Thanks for the feed back. I agree with you about improvements being in the eye of the beholder.
Actually, your frame grab gives me new food for thought i.e; lowering the contrast. I have tweaked the color but not the lightness/darkness.

Will try some more tests when I get home.

Thanks!

--dES

[Des]

Oh, one quick question (I think):
How is Temporal Noise Reduction different from General Noise Reduction, or are the the same thing?
OK that's 2 question.

--dES

[DB83]

Someone may correct me on this.

General noise reduction will attempt to remove noise from the whole frame even if there is no noise. Overdo it and you soften the image too much.

Temporal noise reduction analysises the frame and just removes the noise from the part of the frame that contains the noise and leaves the rest of the frame intact.

[themaster1]

Can you convert the raw avi to huffyuv, a codec most of people use (unlike lagarith) if you still want help..

[Des]

Can you convert the raw avi to huffyuv, a codec most of people use (unlike lagarith) if you still want help..

No problem, here are the same two 10 second files in Huffy.
RAW
http://www.zumodrive.com/share/7f1xOTk4Yj

Filtered
http://www.zumodrive.com/share/7f1KMDMxMW

Thanks.

--dES

[Khaver]

Temporal noise reduction uses a process of comparing single pixels from several frames before and/or after the current frame (temporal=time). It then gathers an average value of all those pixels and if it's different than the current frames pixel by a certain threshold, it changes the current pixel value. The idea being that noise happens on a frame by frame basis and will be different between frames, kind of like lightening doesn't hit in the same place more than once. If you compare a pixel on five frames, as long as there's not a lot of movement, if one pixel is different than the others, it's more than likely noise and can be changed to match the other four pixels.

[Des]

Played around with the filters in Virtualdub some more and got the video to where I am more comfortable with the 'quality' and look of it.
This time the filter chain is: IVTC, DotCrawl, Contrast lowered 13% (thanks DB83 for showing me the effects of that on this video) and Color Equalizer but with a hint of green thrown back in to correct the magenta that was creeping in. Here is the result.

In my head I am hoping for a dramatic improvement, but I realize that the source is just crummy; dirty, unsharp and off color.
What I have managed to do (at least to my eye) is clean up the a good deal of the noise, soften the pallet and correct the color to where the skies are not yellow and the whites are no longer green.

I'll check it out on my TV tonight, and see if it's time to put it rest. Thanks for following along.

--dES

[jagabo]

You need to start with recapturing the video. You have crushed blacks and blown out whites:



In all likelihood there were excursions below y=16 (note the flat bottoms on a graph) and y>235 (note the flat tops) and your capture process cut them off (to keep all pixels in the valid 16<y<235 region.

The best way to adjust your capture levels is to use a capture graph and pipeline the video to an AviSynth script to the VideoScope() or Histogram() filters. You can then use the capture devices proc amp to adjust levels with immediate visual feedback. I've done this with my AVI 650 USB2.

Simple 2d (intra-frame) noise filters average pixel to reduce noise. This results in blurring of the entire image if it's done blindly over the entire image. Smarter ones are "edge sensitive", they don't average at sharp, high contrast, edges so as not to blur them. This gives better results but you will notice that there is more noise near those edges if you look closely. 3D (temporal) noise filters will also look at prior and later frames. Too much "blind" temporal filtering will result in a multiple exposure look (unfortunately, many cable and satellite broadcasts suffer from this). Again, the smarter temporal filters will not average when there is a large difference between the corresponding pixel in the other frames (the assumption being that something moved, or it's a scene change, so by not averaging those pixels you will not get a multiple exposure). The smartest ones also compensate for motion between frames -- in a panning shot for example, they will average with other frames shifted to account for the panning.

[Des]

Thanks Jagabo, I've been capturing straight to the ATI 650 figuring that if I am not change any of the card settings that things are just passing untouched. Clearly now that's not the case.

The best way to adjust your capture levels is to use a capture graph and pipeline the video to an AviSynth script to the VideoScope() or Histogram() filters. You can then use the capture devices proc amp to adjust levels with immediate visual feedback. I've done this with my AVI 650 USB2.

Will I still be using virtualdub for the capture from the card or is this done via another method to pipeline out to the histogram filter?

I'm not unhappy with the results I've gotten so far, but I want to do this correctly and obviously I haven't captured all of the details that are on the tape. Guess I need to go back to the drawing board

Now I'm sorry I squandered my vacation time a few weeks ago by cleaning up the basement and building myself a new office down there instead of learning more about avisynth. I have some time coming up after next weekend to immerse myself in it, meanwhile I will begin reading tutorials and more info regarding avisynth.

Thanks again!

--dES

[jagabo]

The best way to adjust your capture levels is to use a capture graph and pipeline the video to an AviSynth script to the VideoScope() or Histogram() filters. You can then use the capture devices proc amp to adjust levels with immediate visual feedback. I've done this with my AVI 650 USB2.

Will I still be using virtualdub for the capture from the card or is this done via another method to pipeline out to the histogram filter?

You would use the histogram filter just to adjust the proc amp settings of the ATI 650. It looks like I was wrong about using a graphedit filter to capture from the 650 and pipeline to AviSynth. I was unable to duplicate that last night. The problem is that the ATI capture filter has several outputs and importing the filter graph into AviSynth with DirectShowSource() requires that only one output be available. Normally you can route the other outputs to the null renderer in the graph but some of the output pins on the ATI capture filter aren't accepted by the null renderer:



Output2 is the video output. Output3 and Output4 are routed to null renderers (one is audio, the other MPEG audio and video). Output5 and Output6 can't be routed to the null renderer (I don't know what they are). DirectShowSource("cap.grf") complains about there being more than one output pin in the graph.

Another way is to use the Histogram filter in VirtualDub while previewing. Adjust the proc amp so that very little of the video falls in the red ends of the histogram. You can adjust the proc amp with GraphEdit while watching the histrogram in VirtualDub. In GraphEdit insert the ATI AVStream Analog Capture filter. Back at the main GraphEdit window right click on the filter and go to the Video Proc Amp tab. You'll see the histogram in VirtualDub change as you adjust the Proc amp settings. You can access those same settings from within VirtualDub (Video -> Capture Filter) but the histogram doesn't display during the adjustments.

I'm not unhappy with the results I've gotten so far, but I want to do this correctly and obviously I haven't captured all of the details that are on the tape. Guess I need to go back to the drawing board

Your levels aren't too far off. I think you will be able to capture a little more dark detail in the earlier sample though.

[Lodovico]

Another way is to use the Histogram filter in VirtualDub while previewing. Adjust the proc amp so that very little of the video falls in the red ends of the histogram. You can adjust the proc amp with GraphEdit while watching the histrogram in VirtualDub. In GraphEdit insert the ATI AVStream Analog Capture filter. Back at the main GraphEdit window right click on the filter and go to the Video Proc Amp tab. You'll see the histogram in VirtualDub change as you adjust the Proc amp settings. You can access those same settings from within VirtualDub (Video -> Capture Filter) but the histogram doesn't display during the adjustments.

In the VirtualDub Capture window, there is also the Video->Levels control which can make live adjustments to the Proc Amp settings in the preview window and the histogram - if your capture card & driver support it (not all do).

It's a good idea, I think, to crop off any thin black borders from the image before adjusting levels with the histogram. The borders may be darker than any blacks in the rest of the image, and it doesn't matter if they fall into the red zone of the histogram and get crushed. Remember to uncrop after setting the levels and before capturing.

[jagabo]

In the VirtualDub Capture window, there is also the Video->Levels control which can make live adjustments to the Proc Amp settings in the preview window and the histogram - if your capture card & driver support it (not all do).

Unfortunately, that feature doesn't work with the ATI 650 USB2.

It's a good idea, I think, to crop off any thin black borders from the image before adjusting levels with the histogram. The borders may be darker than any blacks in the rest of the image, and it doesn't matter if they fall into the red zone of the histogram and get crushed. Remember to uncrop after setting the levels and before capturing.

And with large black borders (letterboxed movies) the big black peak makes it hard to see the rest of the historgram.

The ColorTools filter in Wave Form Monitor mode works well too. You have to enable the Filter Chain in VirtualDub then add ColorTools:

A levels video cap (keep in mind that VirtualDub performs a rec.601 luma expansion so superblacks and superwhites are crushed):


And the ColorTools waveform (before the rec.601 luma expansion):


The waveform graph has dashed lines at Y=16 and Y=235 but you can't see them in this picture because the bright white lines (from the background) are right on top of them.

[Des]

I started playing with the capture settings in Virtualdub last night, and have discovered as Jagabo mentions, that the levels command is not available to the ati 650. Also the process amp settings for the card are grayed out as well.
The histogram does come on and is active if I have overlay off and preview on, but I cannot seem to adjust the capture level.
Extend black range and extend white range does 'something' in the histogram but I can't see the difference in the video. I am guessing they just give more latitude to the histogram.

This is the 650 pci-e version, but I doubt it's really any different than the usb version.

--dES

[jagabo]

I started playing with the capture settings in Virtualdub last night, and have discovered as Jagabo mentions, that the levels command is not available to the ati 650. Also the process amp settings for the card are grayed out as well.

So when you select Video -> Capture Filter, then select the Video Proc Amp tab you can't move the sliders?

The histogram does come on and is active if I have overlay off and preview on, but I cannot seem to adjust the capture level.

Have you tried using GraphEdit, adding the ATI capture device, and adjusting the proc amp there? (The images in my previous post were made using the ATI 650 USB2, adjusting the levels in GraphEdit.)

Extend black range and extend white range does 'something' in the histogram but I can't see the difference in the video. I am guessing they just give more latitude to the histogram.

You want to leave those disabled.

[Des]

So when you select Video -> Capture Filter, then select the Video Proc Amp tab you can't move the sliders?

Correct. The only slider that seems to be available is the Brightness, which is fine, except that the slider will not move. All the other options are grayed.

I am thinking of using the brightness control on the AVT-8710 to adjust the levels before they come into the ati 650. Do you think this might have a positive effect or just bugger things up more? Right now it is in it's 'neutral' setting.

Have you tried using GraphEdit, adding the ATI capture device, and adjusting the proc amp there? (The images in my previous post were made using the ATI 650 USB2, adjusting the levels in GraphEdit.)

I will try that this evening. I used to use graphedit a long time ago, will need to play a little to refresh my memory.

--dES

[jagabo]

I will try that this evening. I used to use graphedit a long time ago, will need to play a little to refresh my memory.

For this purpose it's very easy. Just start GraphEdit, Select Graph -> Insert Filters, expand the Video Capture Sources tree, and add the ATI analog video capture device. Back at GraphEdit's main window right click on the capture filter and select properties, then go to the Video Proc Amp tab...

[Lodovico]

Correct. The only slider that seems to be available is the Brightness, which is fine, except that the slider will not move. All the other options are grayed.

Are you using Windows 7, by any chance? I found that the Proc Amp controls of all my capture devices stopped working when I installed Windows 7. I now capture on a separate partition running XP.

Extend black range and extend white range does 'something' in the histogram but I can't see the difference in the video. I am guessing they just give more latitude to the histogram.

Extending the luma range during capture forces VDub to scale 0-255 to 16-235. This is useful if you are unable to control the luma range in any other way and you have peaks in the red zones of the histogram which would otherwise be crushed.

[Des]

I managed to use graphedit to create the ...graph?
But I still have no control of the processing amp:



Lodovico wrote:

Are you using Windows 7, by any chance? I found that the Proc Amp controls of all my capture devices stopped working when I installed Windows 7. I now capture on a separate partition running XP.

Yes, I am using W7 along with the 'latest' Ati drivers for W7. I was wondering if having Aero turned on could be the problem, but disabling it seems to have no effect.
Do you dual boot or with XP, or run it in a virtual machine? I would prefer a VM because I'm lazy But I could set up a dual boot.

Extending the luma range during capture forces VDub to scale 0-255 to 16-235. This is useful if you are unable to control the luma range in any other way and you have peaks in the red zones of the histogram which would otherwise be crushed.

So this would 'help' in my current situation?

Funny thing, this video has quickly become more important than it really is in many ways, but considering the learning I am doing and lessons I am receiving here I consider that a good thing! Although I am using the scene I keep posting as a life lesson to myself.

--dES

[Lodovico]

Do you dual boot or with XP, or run it in a virtual machine? I would prefer a VM because I'm lazy But I could set up a dual boot.

I set up a dual boot because I intended to use XP mainly for video work and I wanted it to run as fast as possible. XP in a VM may be fast enough for capture. It depends on your setup.

Extending the luma range during capture forces VDub to scale 0-255 to 16-235. This is useful if you are unable to control the luma range in any other way and you have peaks in the red zones of the histogram which would otherwise be crushed.

So this would 'help' in my current situation?

Yes, it could be a way to get some more detail in the whites and blacks, if you feel that's a problem throughout the capture, and you can't get the Proc Amp controls to work.

[jagabo]

I tried my ATI 650 USB2 on a Win7 computer. I didn't install drivers from the ATI CD. Win7 found drivers when the device was plugged in. Like with your setup, the video proc amp settings aren't functional. That computer is set up with XP Mode for running a few other legacy apps. So I tried using the 650 in XP Mode. I couldn't get XP Mode to see the device though. That's not surprising given XP Mode's paucity of hardware support. Maybe one of the other VMs like VMWare or VirtualBox would work.

<edit>

I have an Ubuntu linux box running Windows XP Pro under VMWare. After installing the driver from the original 650 CD I was able to see the ATI 650 USB2 from VirtualDub and adjust the proc amp settings. VirtualDub crashed when I tried to capture though. Maybe there's some combination of settings that will work...

That computer also has VirtualBox installed. I couldn't get XP under VirtualBox to recognize the ATI 650.

[dphirschler]

You need to start with recapturing the video. You have crushed blacks and blown out whites

Try capturing without the AVT-8710. I've noticed it tends to blow out the whites.


Darryl

[Des]

Darry

Try capturing without the AVT-8710. I've noticed it tends to blow out the whites.

Unfortunately I can't the AVT-8710 corrects keeps the ati 650 from flagging the macrovision.

I need to take a day or two off from this to do 'real work' and rethink my set up.

--dES

[dphirschler]

Damn you, Macrovision!


Darryl