How can I improve this video

[gil900]

OK thanks for the anser.

i there a way to add some opacity on the Right side of the mask that Ghostbuster or exorcist.vdf create?
or Another option is a way to blur the mask.. because if the mask is as it is now so there is no Chance that the mask will cover the Ghost correctly.
the mask is sharp and not like the Ghosts. this is why the mask is visible..

It is indeed reduced the Ghosts, but the mask is visible

[sanlyn]

It's very difficult to judge from the still images. Some of the disturbance looks like halo, some looks like oversharp artifacts, some is ghosting, and some of it is motion smearing and motion trails. You need more than a ghost filter. BHut working only with a single frame, it's impossible to advise.

[gil900]

It's very difficult to judge from the still images. Some of the disturbance looks like halo, some looks like oversharp artifacts, some is ghosting, and some of it is motion smearing and motion trails. You need more than a ghost filter. BHut working only with a single frame, it's impossible to advise.

ok and i know that it is need more filters.

i uploaded tow exampels..
https://forum.videohelp.com/attachment.php?attachmentid=16122&stc=1&d=1359983144
https://forum.videohelp.com/attachment.php?attachmentid=16123&stc=1&d=1359983144

i will very Thank you if you will give the Solution for the ghosts

EDIT:

you don't have try to slove other problams..
just the ghosts.

i think i can deal alone with all other problems

[sanlyn]

Working on it now......

[gil900]

Working on it now......

Thank you!

Most important is that I will understand your solution

[sanlyn]

Well....I don't even know where to start. There are many solutions. Other members might have more ideas. If I kn ew the MaskTools plugin more thoroughly, this might even be easier. I'm still working on this.

When you have a difficult, old, damaged video like these 2 samples, you have to pick a few frames, and start working with one filter and one idea at a time. It takes a while to get some insight this. There is no single filter, no single method. This is what you do: you play -- in video, that's another word for "work". VHS is a "dirty" medium. As tape ages, everything gets corrupt. The magnetic layer becomes altered, even if you don't play the tape. VHS is a terrible medsium for dark areas and wide expanses of flat color. The darkest parts of VHS are filled with noise. In brighter areas you often don't see this noise, but in dark areas they rise up from underground and won't go away.

Note that the images posted here will appear darker in your web browser than they really are. I didn't try to brighten them. You can click on the images and save the original to your hard drive, then view them in a graphics program or VirtualDub.

Below is frame 62 of your dark sample avi. I removed the black borders to prevent them from affecting the histogram. I use RGB histograms to see what the video will look like when displayed in RGB. Others use different methods. The background might appear to be black or dark gray, but it isn't. There is mottling, grain, tape noise, chroma noise, ghosting, chroma bleed, etc.

You said earlier that you "see listen with your eyes". A nice phrase (I like it). Tuning of the eyes comes with practice. The first thing I noticed was not the ghosts -- I would expect problems in dark, flat areas with VHS. No surprise. But the first thing I "heard" with my eyes was serious over saturation. I realize the costumes are supposed to be bright, but these colors are blooming and jumping off the screen. You can see that brights are blown away. Part of the bleeding is due to over saturation.


[Attachment 16124 - Click to enlarge]

One technique often used for this kind of correction is to bring the levels of the background and the noise to the same level. The GhostBuster and filters like it work only with luma, not with chroma. So you can use filters and many tricks to make the background black (below, it is RGB 16), then other filters to reduce the noisy colors and the ghosts to the same level, then yet another filter to brighten data that is below RGB 16 and at the same time to raises invalid colors that are below RGB 16. Keep in mind that TV can't display color below RGB 16 -- it will just be black on TV, period. Then I used other filters to clean some edges, reduce bleed and chroma shift, etc., and to correct the color balance (it was too green), and gamma.

[Attachment 16125 - Click to enlarge]

The b brighter parts of the video have problems that are somewhat worse. Those scenes are over saturated, too bright, and too contrasty. The image below is from frame 83. A histogram of this scene is marked with arrows showing how the colors and luma are clipped (at the right-hand side of the histograms). Blue is so over saturated that even the people are blue, and the brights are clipped to the point where parts of the green and yellow skirts have turned white.

[Attachment 16126 - Click to enlarge]

Contrast and saturation can be altered during capture. They can be altered after capture as well, but if colors/luma are crushed and clipped during capture, they can't often be corrected. Also note that some of the "ghosting" is not ghosting, but shadows on the blue background from some of the dancing figures.

It is all but impossible to correct some of these problems because the capture itself exceeds the acceptable range of video contrast and colors. Correcting this during capture isn't difficult; most capture software and graphics drivers have built-in contrast and saturation filters that can be active during capture. I don't have your capture device or your drivers or software, so I can't advise in detail. I capture with VirtualDub, which allows me to make basic contrast/brightness/saturation corrections during capture.

I used several Avisynth and VirtualDub plugins to correct the dark scenes. The bright scenes will be easier. For this dark scene, I wouild have used NeatVideo to eli9minate a few of the plugins, but since you don't have that filter I had to go with MCTemporalDenoise at "low" to clean up the dirty junk in dark areas. But tape with problems like these takes time and some learning. I don 't know that you want to get that involved. Tapes like this one are not easy.

[sanlyn]

A PAL .m2v of the sample.avi with dark shots, attached. The filters for the brighter parts are not adjusted yet.

[gil900]

Wow thank you!
About how you change the colors, I just (as I said). listen to my eyes.
but i playing with colors through ColorMil and same more plugings in VB .. I do not know exactly what say these graphs. But I just use ColorMil and see what happens.

If there is a color I see it too hard, so I lower the saturation of the color by this for example:



it much easy to do..

Maybe it comes out as good but I do not believe it ..
I think it's a good fast way to change specific colors

[sanlyn]

It's good that you're giving these filters an audition. The only way to learn is to do what all of us do: use them until you drop. But you agould learn to read and work with histograms of some kind, or you're merely trying to outguess yourself.

In any case, with some messy scenes and bad VHS color it's really difficult:

frame 94 original (too "hot", too blue, bleeding red, the "blacks" are blue and so are the figures):

[Attachment 16129 - Click to enlarge]

another attempt. Better, I'm not satisfied. Oh, well...try again....one step at a time.

[Attachment 16130 - Click to enlarge]

But look at the bright dresses, and you can see how clipping has destroyed most of the detail.

[themaster1]

How to apply a filter to dark parts of an image:
http://doom10.org/index.php?topic=2370.0

This will be handy to you i'm sure

[sanlyn]

@master1, I took note of that Doom10 post. Saw it on Doom9 a while back. Haven't had a chance with it yet. I'll know MaskTools one of these days. Eventually.

@gil900:

A): Below, Frame 62 (unmodified original). The RGB histogram shows dark values (left side) and brights (right side). The two thin vertical bars mark RGB 16 (left), RGB 235 (right). The brightness (white area at top) is within the desired RGB 16-240 range for PAL video. On the left side, dark values for all 3 colors are also within the low limit. But on the right side, all 3 colors exceed the brightness range; they extend past RGB 235/240, and even past RGB 255 -- they start climbing the right-hand wall of the histogram. This indicates over saturation and bright clipping. Your eyes can see the effects of high saturation + clipping: loss of bright detail, a glow or "bloom" effect, bleeding, color shift.

[Attachment 16137 - Click to enlarge]

On the left-hand side, the darkest and blackest colors should be at RGB 16 (On TV, RGB 16 displays as "black"). But luma and all colors sit at about RGB 30 (very dark gray). So luma + colors are shifted to the right (too bright). We would say that the black levels are too high, and the white levels as well. Another effect of high brightness and saturation is often an exaggerrated ghosting, which is often present in many analog TV and VHS signals. So you can see three elements of these effects as halos, bleeding, and ghosts. The hazy ghost to the man's right are diffuse image trails.

[Attachment 16138 - Click to enlarge]

The high, sharp peaks at the left-hand side show crushed darks. Crushed data has no dark detail; all of it is more or less a single color or single shade of dark colors, without enough low-end RGB values to reveal fine detail. Brightening doesn't reveal lost values; it merely makes crushed darks "brighter".

B) ColorYUV and Tweak used to adjust levels and saturation. It doesn't look wonderful yet (this won't be the final product). Some bright detail is retrieved and the black levels are closer to normal. It's also easier to see the nature of the bleed, ghosting, etc. You begin to see that the ghosts have "color", and from the skin tones you see that the color balance is too green. By reducing saturation, chroma bleed+shift is less apparent.

[Attachment 16139 - Click to enlarge]

C) Apply 2 excorcist filters. Two are used: one 4% dark mask 42 pixels to the left, and one lighter 2% mask at 28 pixels left. This is imperfect (the images will not display at proper brightness in a web browser), as there is still some "lighter black" in the background at about RGB 24. We know the pixel values of the background noise to be around RGB 15 to 28, because those pixel areas can be read with a CSamp pixel sampler. The CSamp readout is in the right-hand image; a gray arrow marks where readings were made. Not seen in the image is some RGB 0 black mottling in the left part of the screen.

[Attachment 16139 - Click to enlarge]

D) Coring and Gradation Curves added. The coring filters helps smooth black mottling by raising pixel values below RGB 16 to a higher RGB 16. The gradation curve was used to set pixels in the RGB 25 region to RGB 16, and to tweak the midtones a other areas. The background is now uniform, and exorcist is difficult to see. Again, not perfect (the guy is a little blurry, but he's in fast motion here).

[Attachment 16140 - Click to enlarge]

E (left): Add more AviSynth Cleanup. F (right): More VirtualDub tweaks. Tweak to your heart's content.

[Attachment 16141 - Click to enlarge]

Some avisynth script and VDub notes, functions, and VirtualDub .vcf file attached.

[gil900]

Thank you very much sanlyn!

I'll use it and learn what you did

[sanlyn]

Hope you found some ideas to play with. I really hate ghosts in big, flat, dark areas, they can drive you crazy. Will tackle the easier ghost problems later tonight.

[gil900]

Yes.
You gave me a good idea and I understood it.

But your script is very hard to CPU.
the Processing speed is 0.57FPS.

very slow speed.
my goal is about 3FPS+ Processing speed.

but i got the main Idea and i built new script with the idia that the Processing speed is 3-6 FPS (Includes important filters)

I think you treat it like it's a picture .. Do not forget that this is a video should go Processing ..

i will test what themaster1 gave (thanks to themaster1)

[sanlyn]

Yes.
You gave me a good idea and I understood it.

But your script is very hard to CPU.
the Processing speed is 0.57FPS.

very slow speed.
my goal is about 3FPS+ Processing speed.

The slow plugin is MCTEmporalDenoise, even at "Low", but it cleans up the tape noise and motion shimmer. I would have suggested replacing MCTemporalDenoise with NeatVideo as a later step in VirtualDub, but you don't have that filter. Also, you can probably eliminate DeBlock(), but you still need GradFun2DBmod for the coarse grain and hard edges in shadow areas.

Replacing MCTD with NeatVideo, the script ran about 2.5 fps, then the VirtualDub step ran 12 fps with NeatVideo. Many users can't live without NeatVideo for old, aged, noisy tape -- but you have to learn to tweak NeatVideo, and you can't use it for everything in sight.

but i got the main Idea and i built new script with the idia that the Processing speed is 3-6 FPS (Includes important filters)

I think you treat it like it's a picture .. Do not forget that this is a video should go Processing ..

i will test what themaster1 gave (thanks to themaster1)

Video is a stream of "pictures". The principles for noise control, masking, color correction, etc., are the same for photo and video alike. Most of the filters used in the script are spatio-temporal, temporal-only, or motion-compensated filters, which won't work on still images. IF you look at a video one frame at a time, you will not see most of the noise; but in video the noise varies from image to image, and thus becomes more visible.

There is more than one way to accomplish most tasks. You are correct -- the way to get an idea about what various techniques will accomplish is to use them and see what happens.

[sanlyn]

Spent some time with other scenes in your sample.avi and sample 2 .avi. The captures are too bright, too contrasty, and seriously over saturated. Bright areas are clipped so badly in some scenes that the brightest highlights actually change color. This level of contrast and oversaturation makes tv broadcast ghosting look worse and causes much of the bad chroma bleeding. Chroima bleed also covers detail, which often looks smeared when the bleed and color shift are processed.

In some cases, darkening clipped highlights reveals the loss of detail and hue distortion.

[Attachment 16170 - Click to enlarge]

Attached, a combined mkv of joined sample.avi and sample 2.avi after some corrections. Because of the problems described above, some aspects are beyond repair. The scene with spotlights in the attached mkv are another example of high saturation and contrast causing bleed and color change. It's not difficult to control contrast and saturation during capture.

[gil900]

Thank you!

Already done with this recording ..
I got a similar result to what you got.

It's not what matter ...

I have a new example that I think you will be shocked the amount of noise there was.
I'll send it to you tomorrow ..

I have to use MCTD on high settings + NeatVideo on high settings + super temporary cleaner 1.4 MT to cleen all the noise ..

i have a picture for you:


I also use
ChromaShift(C=2,L=-2)
mergechroma(aWarpSharp(depth=50))

for the big Chroma problem (as you can see on the last image)

it takes 1.25 days to Process this

[sanlyn]

Biggest problem: according to the images you have posted, you could have saved much time and trouble by lowering saturation and contrast during capture. None of the filters you described will address those defects.

[gil900]

ן always lower the saturation in VirtualDub.

i dont know how to do this in debut video capture

[sanlyn]

By the time the video gets into VirtuaDub, it's too late to correct most of the Contrast and Sat levels. I don't have your capture software, but you showed menus that set several options. Surely there is a basic Bright, Contrast, and Saturation (or "Intensity") setting. These are the only filters that you should use during capture, anyway -- denoisers are too slow, and hue controls never give what the video needs. I capture with VirtualDub Capture, and that utility allows me to access the Bright/Contrast/Sat filters in my ATI capture card.

[sanlyn]

ן always lower the saturation in VirtualDub.

i dont know how to do this in debut video capture

The NCH website says that Debut has adjustments for brightness, contrast, and gamma. Start with Contrast first. http://www.nchsoftware.com/capture/index.html


[Attachment 16183 - Click to enlarge]

If you can't recapture, your other recourse is to adjust contrast, levels, etc., before converting to RGB and going to VirtualDub. We've covered that subject previously. At least preparing with ColorYUV and/or Tweak in YUV will help prevent more damage with the RGB conversion of non-standard luma and chroma values.

[jagabo]

VirtualDub's Brightness/Contrast filter can work in YUV. So just apply that filter first (if you can't re-capture).

[gil900]

thanks sanlyn.
i found how to do this in Debut Video Capture Software.

i wanted to say that the Debut Video Capture Software is the best softwer that i found on the internet.
and i found that im not the only one who say that.. it is relley good softwer.
every other softwer didnt work well with my Capture device.
VirtualDub is the software that work worst for this.. not stible with my device and have meny bugs..
the same about other softwers.

only Debut work prefectly.

Until now, I had no problems with the software.
Before I found this program, I recorded on DVD. And I did not use my device for that because of problems ..

I'm surprised you did not know this software ..

But here there anyone else who thinks so:
http://www.engon.de/audio/vhs0_en.htm

and he have the same device i have

[sanlyn]

I don't need extra software. VirtualDub does not recognize some capture devices, so that's no surprise. I capture VHS with VirtualDub using ATI All-In-Wonder Radeon 7500 and 9600XT capture cards.