How can I improve this video

[gil900]

Hello,
I finished record and process a tape ..
And deleted the original recording.

I used some small color changes in VirtualDub and Noise reduction and nothing more ..

I I attached the finished video file.

What needs to improve on?
I think there is a problem with sharpness ..

[jagabo]

Don't deinterlace or deinterlace better (depending on your intended use). Restore the original frame rate (get rid of duplicate frames). Levels and colors look about right. I saw a lot of macroblock artifacts. Did you capture with MPEG or MJPEG?

[gil900]

Ok

How do I fix this:



I have no idea how to fix it ..

I have tapes with color leakage also that I do not know the fix.

[jagabo]

Those are oversharpening halos, probably caused by the VHS deck. You can try using de-halo filters in AviSynth. Or use a deck without sharpening filters. In the end you'll be left with fuzzier video though.

[gil900]

Those are oversharpening halos, probably caused by the VHS deck. You can try using de-halo filters in AviSynth. Or use a deck without sharpening filters. In the end you'll be left with fuzzier video though.

thanks..

like this:
MT("BlindDeHalo3(rx=2.0,ry=2.0,strength=63)",threa ds=2)

?

I know this is the tape machine

i set their the sharpnesses on medium.

What do you recommend me?
to set in the tape machine the lowest sharpness?

[jagabo]

Turn off the sharpness filter on the deck. VHS halos are only on the x dimension. Use a larger x than y in dehalo filters. Around 4 in that sample. You'll also have to crank up the strength.

[sanlyn]

Hello,
I finished record and process a tape ..

That doesn't tell us much. How did you record? What player? what capture device? What capture format? What does "processing" mean? And so forth.

And deleted the original recording.

I used some small color changes in VirtualDub and Noise reduction and nothing more ..

No, you did more than that. The video is deinterlaced.

What needs to improve on?
I think there is a problem with sharpness ..

Can't answer that question without more info. But as jagabo has suggested, sharpness isn't the only problem. BTW, if you intend to go with DVD on this video, your frame rate and audio sample rate are not compliant for either DVD or Standard Definition BluRay or AVCHD. But that depends on what you want for final output.

[gil900]

Really a disaster that deleted the original recording ..
But it's not such a big disaster ..

I can record again.

The problem is that I have over 30 tapes to work on them ..


about the de-halo filters, this is work for me:
FFT3DFilter(last, bt=-1, plane=0, dehalo=4.0, hr=2.0, ht=50, svr=0)

[sanlyn]

My guess about your deinterlace procedure is that you used a simple bob() or yadif. The former seems more likely because this video looks soft, and bob doesn't produce the best deinterlacing. A very low h264 bitrate of 1146 kbps diesn't help.

As far as I can recall, your previous video posts were PAL videos. How did the frame rate of this posted sample get to be 30.068 FPS ? ? The sample also appears to have been resized, but I could be incorrect there.

[gil900]

My guess about your deinterlace procedure is that you used a simple bob() or yadif. The former seems more likely because this video looks soft, and bob doesn't produce the best deinterlacing. A very low h264 bitrate of 1146 kbps diesn't help.

As far as I can recall, your previous video posts were PAL videos. How did the frame rate of this posted sample get to be 30.068 FPS ? ? The sample also appears to have been resized, but I could be incorrect there.

How did the frame rate of this posted sample get to be 30.068 FPS ? ? The sample also appears to have been resized, but I could be incorrect there.

I used edited Audio file that filtered from noise
I did the audio file with the software: WavePad Sound Editor

I had to change the FPS of the video from 30.00 to FPS of the new audio file.

Otherwise the audio was not appropriate for video

My guess about your deinterlace procedure is that you used a simple bob() or yadif. The former seems more likely because this video looks soft, and bob doesn't produce the best deinterlacing. A very low h264 bitrate of 1146 kbps diesn't help.

i did the deinterlace in VirtualDub..

i can do it witch this:
Import("C:\Program Files (x86)\AviSynth 2.5\plugins\DeVCR.avs")
DeVCR(20)

SelectEvery(4,2,3)


Is that better?


After i did the "dehalo=4.0",
This new problem:


I do not know if it's because "dehalo=4.0"
(the same tape and show Containing a anime..)

But how do I fix it?

[sanlyn]

I had to change the FPS of the video from 30.00 to FPS of the new audio file.

Otherwise the audio was not appropriate for video

IF your intention is DVD or BD/AVCHD, it's still not appropriate.

i did the deinterlace in VirtualDub..

That could explain some of the problems, and unmanaged colorspace conversion likely added to it.

i can do it witch this:
Import("C:\Program Files (x86)\AviSynth 2.5\plugins\DeVCR.avs")
DeVCR(20)

SelectEvery(4,2,3)


Is that better?

DeVCR should not be used on interlaced video, and "SelectEvery()" is incorrectly applied. Use QTGMC for best results. DeVCR can't repair chroma and luma flashing.

After i did the "dehalo=4.0",
This new problem:


[Attachment 16038 - Click to enlarge]
I do not know if it's because "dehalo=4.0"
(the same tape and show Containing a anime..)

But how do I fix it?

You need to recapture and work with the capture first. The cartoon image is blurry, has aliasing and other problems, and is seriously over-saturated.

[gil900]

You need to recapture and work with the capture first.

Yes that is exactly what I was going to do.

I'll come back after I finish ..

Thank you

[sanlyn]

Good idea. It will save you a lot of endless work and corrections. In any case, for the most part the color in your original posted AVI is generally good, with just a few problems.

After you capture, members can make more accurate assessments if you post a short sample of your capture, unprocessed. And let us know what you desire for your final output: DVD ? Computer playback? What you want for final output will determine most of your processing.

[gil900]

Ok I finished recording and I I've attached a new sample file.

First time I write code myself.
This is what I wrote:

loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\FFT3DFilter.dll")
Import("C:\Program Files (x86)\AviSynth 2.5\plugins\QTGMC-3.32.avs")

last = Mpeg2Source("1.d2v")

last = MergeChroma(last, McTemporalDenoise(settings="very high"))
FFT3DFilter(last, bt=-1, plane=0, dehalo=2, hr=3.0, ht=50, svr=0)

QTGMC( Preset="Slower", Sharpness=0.75 )
SelectEven()

return last

But there is a problem with the colors .. They smeared ..
I do not know how to fix it.

IF your intention is DVD or BD/AVCHD, it's still not appropriate.

I checked and it works on the Xtreamer ..
But I prefer it to work on as many devices.

Better with 30FPS

[sanlyn]

The m2v has been processed. We need a sample of the original, unprocessed capture (If you destroyed the original again, you will be shot at sunrise ).

Why are you deinterlacing?

But there is a problem with the colors .. They smeared ..
I do not know how to fix it.

Unable to advise. Because this is not the original capture, can't say why the filters were needed in the first place. In any case, a lot of noise is still there.

The video is smeared by running two plugins that are not designed for interlaced video. Don't know why you ran MCTemporalDenoise at "very high", and there is virtually no difference between running QTGMC at its default "slow" settings and running it at "very slow". In any case, the videos are smeared not only because the plugins are run incorrectly and their settings are too strong, but they might not have been appropriate for the original capture's problems.

Not that I think the script you used was appropriate, but if you had to run these plugins you would run them this way:

loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\FFT3DFilter.dll")
Import("C:\Program Files (x86)\AviSynth 2.5\plugins\QTGMC-3.32.avs")

Mpeg2Source("1.d2v")
MergeChroma(McTemporalDenoise(settings="very high", interlaced=true))
QTGMC( Preset="Slower", Sharpness=0.75 )
SelectEven()
FFT3DFilter(bt=-1, plane=0, dehalo=2, hr=3.0, ht=50, svr=0)

You will likely get better results against halo and chroma bleed with these:
mergechroma(aWarpSharp(depth=20.0, thresh=0.75, blurlevel=3, cm=1))
DeHalo_Alpha()

However, none of those plugins will solve the vertical edge ghosting.

I checked and it works on the Xtreamer ..
But I prefer it to work on as many devices.

Better with 30FPS

Then you will have limited playback capability with this format. There's no way you can encode this to any DVD or AVCHD/BD standard, which would offer wider playback compatibility.

Your whites are blown out, and there's a luminance change between frames 710 and 724. The first problem should be fixed, but most viewers would probably be able to live with the second.

[gil900]

I use McTemporalDenoise filter because small sections with the problem that the filter designed to solve ..
what is interlaced video?

However, none of those plugins will solve the vertical edge ghosting. Which

Which plugin can slove this?


This is the original recording .. I cut a small part with DGIndex.
I recorded with "Debut Video Capture" Software
to a mpg file

[sanlyn]

what is interlaced video?

If you don't know what interlaced video is, why did you use QTGMC?

This is the original recording .. I cut a small part with DGIndex.
I recorded with "Debut Video Capture" Software
to a mpg file

Then I'm afraid we need more in formation about your m2v sample. It has some serious problems. The video is interlaced, but it appears to have been encoded as progressive. Where did this video originate? Is it a tape? Is it something recorded from TV or from a commercial disc? Part of it is TFF (Top Frame First) but there are BFF frames where motion is reversed between fields. Commercial tapes are 25 fps (PAL) or 29.97 fps (NTSC), your m2v sample is neither.
Motion is reversed in one field every 6 frames using a duplicate of a previous field in reversed field order.

I'm under the impression that you included the script in your previous post because you had used it with this video.

If your m2v sample is a cut directly from the original, unprocessed capture, how did it become 30 fps? Was it recorded that way? Is it possible to record 25 fps interlaced source into progressive mpeg at 30 fps with your capture software ?

Try this: Open your m2v sample this way:

MPEG2Source("[enter path to your d2v project]\1.demuxed.d2v")
AssumeTFF().SeparateFields()

Open that script in VirtualDub and use the > frame button to watch the video one frame at a time. Remember that you are looking at fields, not frames. There are two fields in each frame. Advance to field 33, view one field at a time, and you will see that the motion is reversed (a field is duplicated) every 12 fields (or every 6 frames). Are you saying that this is the way your video was recorded?

[gil900]

I did not recorded on DVD ..
I recorded with the easycap.

And these are my settings:

If your m2v sample is a cut directly from the original, unprocessed capture, how did it become 30 fps? Was it recorded that way? Is it possible to record 25 fps interlaced source into progressive mpeg at 30 fps with your capture software ?

yes..
The answer is simple than you think.
It's not

I saw the quality remains the same.. I have convert over 30 tapes.

I have no hard disk have so many recordings of this size(unprocessed) ..

[aedipuss]

you didn't get the easy part right. mpeg2 dvd standard 4:3 ntsc tape is 720x480, pal 720x576.

[sanlyn]

It is


[Attachment 16050 - Click to enlarge]


[Attachment 16051 - Click to enlarge]

Media Info:

General
Complete name : E:\forum\improve\1.demuxed.m2v
Format : MPEG Video
Format version : Version 2
File size : 24.4 MiB
Duration : 32s 533ms
Overall bit rate mode : Variable
Overall bit rate : 6 279 Kbps

Video
Format : MPEG Video
Format version : Version 2
Format profile : Main@Main
Format settings, BVOP : Yes
Format settings, Matrix : Default
Format settings, GOP : M=3, N=12
Duration : 32s 533ms
Bit rate mode : Variable
Bit rate : 6 279 Kbps
Maximum bit rate : 9 000 Kbps
Width : 640 pixels
Height : 480 pixels
Display aspect ratio : 4:3
Frame rate : 30.000 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Compression mode : Lossy
Bits/(Pixel*Frame) : 0.681
Stream size : 24.4 MiB (100%)

The m2v is 965 frames. Playing time is 32 seconds. (966 frames) X (30 fps) = a pinch over 32 seconds.

[gil900]

So in my language -
Possible to fix it?

It's shorter than it should be?


What I have set up in the recording software for it will not happen again?


EDIT:
If the tape is not 30 FPS (PAL), so it is better the standard FPS of the tape ..

[sanlyn]

Standard PAL VHS is 25 FPS, usually interlaced with Top Field First. I have no idea how the recording came to be 30 FPS, unless NTSC somehow came into the picture, but even that wouldn't be correct. The original soutrce from which the tape was made appars to be film-based at 23.79 FPS. Usually, film is adjusted to 25 FPS PAL by inserting one frame every 24 frames, or some other medtod such a weird forms of telecine. It would have been smoother on playback if 23.97 FPS had been telecined with 3:2 pulldown to get NTSC 29.972 FPS, which will play on most European machines.

So, I can't see how this 30 FPS occurred. Obviously an extra "field" has been added, but it is out of sequence and actually belongs to a previous frame. If you use the SeparateFields script I suggested earlier, and go to field 32, then look at fields 32, 33, 34, 35. You will see that field 34 is a "step backward". Actually, field 34 is a duplicate of field 32. This happens every 12 fields (or every 6 frames). Now, how to get rid of that extra field ( or put it into its proper place) is another matter. I can only guess. Some members who have seen this oddity elsewhere might have the answer.

[poisondeathray]

It would be better to get a proper recording, but you can decimate most of the duplicates by using TDecimate.

e.g.
TDecimate(mode=1, cycleR=1, cycle=6)

[sanlyn]

Yes, thank you. I was looking into TDecimate docs but was interrupted when trying to figure how to state the pattern. Will have to try that tomorrow, it's getting late here. Heaven knows what FPS we'll get, but we'll see.

It remains to be seen if the original tape is really set up that way. Oh, well....

ED: Just couldn't wait. Tried a few frames quickly. After TDecimate it's 25FPS, interlaced. Will check more tomorrow <big yawn>.

[gil900]

Ed2: Note, your sample.avi from post #1 has a frame rate of 30.068 FPS. A mystery to me. Both of the sample videos appear to have used ssome form of telecine. You should not deinterlace telecined video.

The first example is Should not be really a 30.068 FPS ..

I already said - I changed the FPS to the FPS of the new audio. I had to match audio to video.

And I'm talking about edited audio. The original audio is not really 30.068..

do you what a sample of the audio file? (the new original recoding)
if it can help you to solve the mystery so i can upload for you.

So, I can't see how this 30 FPS occurred. Obviously an extra "field" has been added, but it is out of sequence and actually belongs to a previous frame. If you use the SeparateFields script I suggested earlier, and go to field 32, then look at fields 32, 33, 34, 35. You will see that field 34 is a "step backward". Actually, field 34 is a duplicate of field 32. This happens every 12 fields (or every 6 frames). Now, how to get rid of that extra field ( or put it into its proper place) is another matter. I can only guess. Some members who have seen this oddity elsewhere might have the answer.

i know what you talking about.. i don't need to see it on the example.. i see it on my full capture..
i really see that every 6 frames there is a duplicate frame.
you have a verry good eyes.

but i dont know why this is "a serious problem".
if i play the video so it playng perfectly..
Do you think it played slower than it should be?

my problem that i want to fix is the color bleeding..(Or whatever it is)
You're right, the recording is not bad. But if I can fix the color bleeding then the video will look like new and i dont want to Give it up.

poisondeathray:
It would be better to get a proper recording, but you can decimate most of the duplicates by using TDecimate.

e.g.
TDecimate(mode=1, cycleR=1, cycle=6)

ok thanks.. if it will slove the problem than it is good.
but what about the audio? Maybe i need to change the FPS of the audio?

EDIT:

It really solved the problem!
But sometimes there are still a few frames remained duplicate.

But it does not happen often .. (I check it on full capture)


EDIT:
the FPS of the video is now 23.976
and i changed the mode in
TDecimate(mode=1, cycleR=1, cycle=6)
to mode 2
now it is:
TDecimate(mode=2, cycleR=1, cycle=6)

I read the document and saw that mode 2 is the best. (Based on good algorithm)

but as i expected, now i have a problem wich the sound.. to match the video to the sound, i need to change the FPS of the video to 16.614

[jagabo]

every 6 frames there is a duplicate frame...
but i dont know why this is "a serious problem".
if i play the video so it playng perfectly...

It's not playing perfectly. There are five jerks in the motion every second. Since this is interlaced video the video plays backwards at each of those jerks.

my problem that i want to fix is the color bleeding..

The color bleeding isn't hard to fix. You can use something like:

MergeChroma(awarpsharp2(depth=30))

after a smart bob deinterlace.

But if your source is PAL you should really get a proper 720x576 25 fps capture first. Worry about the audio sync later. Actually there shouldn't be an issue with the audio sync: a 25 fps source captured at 30 fps (with a duplicate every 6th frame) has the same running time as the original 25 fps. But capturing PAL at 640x480 at 30 fps is doing severe damage to the picture.

Maybe we should verify if you're really capturing PAL video? What country are you in? What is the brand and model number of your VCR?

[gil900]

thnaks jagabo

i will try this and update if it works.

i added a new information in my #25
and I described a new problem with the audio.

i need to solve this now..

Maybe we should verify if you're really capturing PAL video? What country are you in? What is the brand and model number of your VCR?

yes this must be a pal video.. the new FPS i got after the fix is PAL FPS (23.976) and i live in israel..
i my country this is pal.

my VCR model is
TOSHIBA VE35

so you say that if i will capture on 720x576 , 25 FPS so all thows problems will solve?
but why not capture on 23.976?

EDIT:
oops..
i chaked on google and sow that 23.976FPS is NTSC..
this is strange

[jagabo]

i added a new information in my #25
and I described a new problem with the audio.
i need to solve this now..

No, you need to get a proper capture first. Worry about audio sync after you've done that. TDecimate() doesn't change the running time and does not create A/V sync problems. 30 frames played at 30 fps takes 1 second to play. 25 frames played at 25 fps also takes 1 second. There is no change in running time from TDecimate.

i live in israel..
i my country this is pal.

So yes, you should be capturing with PAL settins, not NTSC.

so you say that if i will capture on 720x576 , 25 FPS so all thows problems will solve?

All your problems solved? Probably not. But you will get much cleaner captures.

but why not capture on 23.976?

Because it's screwing up your video. You are getting duplicate frames (not all of which will be fixed correctly by TDecimate), the video is being resized poorly (576 lines to 480 lines), and the chroma channels are getting screwed up.

[gil900]

i added a new information in my #25
and I described a new problem with the audio.
i need to solve this now..

No, you need to get a proper capture first. Worry about audio sync after you've done that. TDecimate() doesn't change the running time and does not create A/V sync problems. 30 frames played at 30 fps takes 1 second to play. 25 frames played at 25 fps also takes 1 second. There is no change in running time from TDecimate.

i live in israel..
i my country this is pal.

So yes, you should be capturing with PAL settins, not NTSC.

so you say that if i will capture on 720x576 , 25 FPS so all thows problems will solve?

All your problems solved? Probably not. But you will get much cleaner captures.

but why not capture on 23.976?

Because it's screwing up your video. You are getting duplicate frames (not all of which will be fixed correctly by TDecimate), the video is being resized poorly (576 lines to 480 lines), and the chroma channels are getting screwed up.

OK i will do what you say
but 23.976 FPS is not NTSC ?

[jagabo]

but 23.976 FPS is not NTSC ?

Analog NTSC video is always 59.94 fields per second. When captured, pairs of fields are combined into frames to make 29.97 frames per second digital video. Film sources are often stored on DVD as 23.976 progressive frames per second along with pulldown flags that instruct the DVD player on how to produce 59.94 fields per second analog output that can be viewed on analog TVs.