Cleaning trash, fps, and some other issues, help.

[johnmeyer]

You can remove a lot of the "comets" with RemoveDirtMC() or RemoveSpotsMC().

Yes, that is true; I have used them both for this purpose.

Two comments:

1. That bad video looks like the result of bad tracking. Since this is MKV, I assume you got it from someone else, but if you have the tape and can do a new transfer, use the manual tracking control. I'll bet you can make it look almost perfect.

2. If you can't re-transfer, then jagabo is correct. However, RemoveSpots requires a LOT of tuning and is not at all easy to use. Therefore, you might want to start with a function I wrote several years ago (see below) and which I just recently used on two projects. It worked absolutely perfectly, removing almost all the horizontal white dropouts. However, my source was a good deal cleaner than yours, and I am quite certain that you will not get results that are as good as what I got. The reason is that you have too many frames where the "dropout" persists for more than one frame. That makes removal quite difficult.

Here's my function. As you can see, it was written for interlaced material, but if your source is progressive (I don't have time to download and test it), this should still work just fine. It is motion compensated, just like RemoveSpotsMC, and might even contain some of the same code (I wrote it so long ago that I no longer remember from who I stole pieces of code.

function Remove_Tears(clip source) {

  #Create mask
  ml           = 100 # mask scale
  scene_change = 400 # scene change

  super = MSuper(source,pel=2, sharp=1)

  vf = MAnalyse(super,isb=false) # forward vectors 
  vb = MAnalyse(super,isb=true)  # backward vectors 

  cf = MFlow(source,super,vf,thSCD1=scene_change) # previous compensated forward
  cb = MFlow(source,super,vb,thSCD1=scene_change) # next compensated backward 

  sadf = MMask(super,vf, ml=100,kind=1,gamma=1, thSCD1 = scene_change)
  msadf= sadf.Binarize() # binary inverted forward SAD mask

  sadb = MMask(super,vb, ml=ml, gamma=1, kind=1, thSCD1 = scene_change) # backward SAD mask
  msadb= sadb.Binarize() # binary inverted backward SAD mask

  msad = Logic(msadf,msadb,"OR") # combined inverted SAD mask
  msad = msad.Expand() # expanded inverted SAD mask
  msadi = Interleave(msad, msad, msad) # interleaved 3-frame inverted SAD mask

  Interleave(cf,source,cb) # interleave forward compensated, source, and backward compensated

  #DeSpot(show=0,p1percent=10,dilate=1,maxpts=400,p2=12,mthres=18,p1=24,pwidth=40,pheight=12,mwidth=7,mheight=5,merode=33,interlaced=false,seg=1,sign=1,ranked=true,extmask=msadi)

  DeSpot(show=0,p1percent=3,dilate=3,maxpts=400,p2=6,mthres=9,p1=12,pwidth=140,pheight=4,mwidth=7,mheight=5,merode=33,interlaced=false,seg=1,sign=-1,ranked=true)

  SelectEvery(3,1)
}

[jagabo]

The grotesque anomaly video is one of the worst I've ever seen. It's worth a download just to see how bad it is.

I tried your remove_tears() function and it didn't much to his video (or another much cleaner video I tried). But I had to replace binarize(), logic(), etc. with the mt_ equivalents. And I don't know if the despot() I was using is the same as yours.

[Caiosouza]

You can remove a lot of the "comets" with RemoveDirtMC() or RemoveSpotsMC().

Yes, that is true; I have used them both for this purpose.

Two comments:

1. That bad video looks like the result of bad tracking. Since this is MKV, I assume you got it from someone else, but if you have the tape and can do a new transfer, use the manual tracking control. I'll bet you can make it look almost perfect.

2. If you can't re-transfer, then jagabo is correct. However, RemoveSpots requires a LOT of tuning and is not at all easy to use. Therefore, you might want to start with a function I wrote several years ago (see below) and which I just recently used on two projects. It worked absolutely perfectly, removing almost all the horizontal white dropouts. However, my source was a good deal cleaner than yours, and I am quite certain that you will not get results that are as good as what I got. The reason is that you have too many frames where the "dropout" persists for more than one frame. That makes removal quite difficult.

Here's my function. As you can see, it was written for interlaced material, but if your source is progressive (I don't have time to download and test it), this should still work just fine. It is motion compensated, just like RemoveSpotsMC, and might even contain some of the same code (I wrote it so long ago that I no longer remember from who I stole pieces of code.

Thank you! I do not have the source, I would never try to capture something like that, and if I did try it the best way possible, then anything, any improvement is welcome on this aberration

The grotesque anomaly video is one of the worst I've ever seen. It's worth a download just to see how bad it is.

I tried your remove_tears() function and it didn't much to his video (or another much cleaner video I tried). But I had to replace binarize(), logic(), etc. with the mt_ equivalents. And I don't know if the despot() I was using is the same as yours.

Good thing that you returned to be in a good mood,I did not throw it away because this is a rare video of this segment of sports, some improvement would be welcome, if you are terrified, imagine me looking at this tangle of drizzles, I do not have the least idea where to start.

The sample are not good,I think the freezes no have on source, I will load a new segment,

If you like challenges this is your.

[johnmeyer]

I should have looked more closely at the "Remove_Tears" function I posted. It is actually designed for progressive footage. This is really important for your sad clip because if you separate that clip into fields, you will find that while there is no motion between fields, the noise bars are quite different between the even and odd fields. Therefore the fields must be cleaned separately from each other and then reassembled.

Here is a crudely-constructed interlaced version of the function. I also found that because there is so much noise and artifacts from the extremely low bitrate encoding used somewhere along the line, that you have to run a general denoiser first. I've also attached two photos showing before and after on one frame. The user needs to play around with the Despot settings to get the rest of the spots. I think that with the proper tuning that it should be able to get almost all of them.

P.S. Please note that in this badly-constructed code, Despot appears in two places, and if you change the settings on one, you have to make the other match exactly. This is easy to do with a copy/paste operation, but you have to remember to do it. Sorry for the bad coding, but I don't have time to do it right.

function Remove_TearsI(clip source) {

#Create mask
ml           = 100 # mask scale
scene_change = 900 # scene change

fields=source.SeparateFields()
even_fields=selecteven(fields)
odd_fields =selectodd(fields)

super_even = MSuper(even_fields,pel=2, sharp=1)
vfe = MAnalyse(super_even,isb=false) # forward vectors 
vbe = MAnalyse(super_even,isb=true)  # backward vectors 

cfe = MFlow(even_fields,super_even,vfe,thSCD1=scene_change) # previous compensated forward
cbe = MFlow(even_fields,super_even,vbe,thSCD1=scene_change) # next compensated backward 

sadfe = MMask(super_even,vfe, ml=100,kind=1,gamma=1, thSCD1 = scene_change)
msadfe= sadfe.Binarize() # binary inverted forward SAD mask

sadbe = MMask(super_even,vbe, ml=ml, gamma=1, kind=1, thSCD1 = scene_change) # backward SAD mask
msadbe= sadbe.Binarize() # binary inverted backward SAD mask

msade = Logic(msadfe,msadbe,"OR") # combined inverted SAD mask
msade = msade.Expand() # expanded inverted SAD mask
msadie = Interleave(msade, msade, msade) # interleaved 3-frame inverted SAD mask

Interleave(cfe,even_fields,cbe) # interleave forward compensated, source, and backward compensated

even_result_fields=DeSpot(show=0,mscene=99,p1percent=3,dilate=3,maxpts=1000,p2=6,mthres=9,p1=12,pwidth=840,pheight=12,mwidth=7,mheight=5,merode=33,interlaced=false,seg=1,sign=0,ranked=true).SelectEvery(3,1)

super_odd = MSuper(odd_fields,pel=2, sharp=1)
vfo = MAnalyse(super_odd,isb=false) # forward vectors 
vbo = MAnalyse(super_odd,isb=true)  # backward vectors 

cfo = MFlow(odd_fields,super_odd,vfo,thSCD1=scene_change) # previous compensated forward
cbo = MFlow(odd_fields,super_odd,vbo,thSCD1=scene_change) # next compensated backward 

sadfo = MMask(super_odd,vfo, ml=100,kind=1,gamma=1, thSCD1 = scene_change)
msadfo= sadfo.Binarize() # binary inverted forward SAD mask

sadbo = MMask(super_odd,vbo, ml=ml, gamma=1, kind=1, thSCD1 = scene_change) # backward SAD mask
msadbo= sadbo.Binarize() # binary inverted backward SAD mask

msado = Logic(msadfo,msadbo,"OR") # combined inverted SAD mask
msado = msado.Expand() # expanded inverted SAD mask
msadio = Interleave(msado, msado, msado) # interleaved 3-frame inverted SAD mask

Interleave(cfo,odd_fields,cbo) # interleave forward compensated, source, and backward compensated

odd_result_fields=DeSpot(show=0,mscene=99,p1percent=3,dilate=3,maxpts=1000,p2=6,mthres=9,p1=12,pwidth=840,pheight=12,mwidth=7,mheight=5,merode=33,interlaced=false,seg=1,sign=0,ranked=true).SelectEvery(3,1)
interleave(even_result_fields,odd_result_fields)
weave()
}

[Caiosouza]

I should have looked more closely at the "Remove_Tears" function I posted. It is actually designed for progressive footage. This is really important for your sad clip because if you separate that clip into fields, you will find that while there is no motion between fields, the noise bars are quite different between the even and odd fields. Therefore the fields must be cleaned separately from each other and then reassembled.

Here is a crudely-constructed interlaced version of the function. I also found that because there is so much noise and artifacts from the extremely low bitrate encoding used somewhere along the line, that you have to run a general denoiser first. I've also attached two photos showing before and after on one frame. The user needs to play around with the Despot settings to get the rest of the spots. I think that with the proper tuning that it should be able to get almost all of them.

P.S. Please note that in this badly-constructed code, Despot appears in two places, and if you change the settings on one, you have to make the other match exactly. This is easy to do with a copy/paste operation, but you have to remember to do it. Sorry for the bad coding, but I don't have time to do it right.

function Remove_TearsI(clip source) {

#Create mask
ml           = 100 # mask scale
scene_change = 900 # scene change

fields=source.SeparateFields()
even_fields=selecteven(fields)
odd_fields =selectodd(fields)

super_even = MSuper(even_fields,pel=2, sharp=1)
vfe = MAnalyse(super_even,isb=false) # forward vectors 
vbe = MAnalyse(super_even,isb=true)  # backward vectors 

cfe = MFlow(even_fields,super_even,vfe,thSCD1=scene_change) # previous compensated forward
cbe = MFlow(even_fields,super_even,vbe,thSCD1=scene_change) # next compensated backward 

sadfe = MMask(super_even,vfe, ml=100,kind=1,gamma=1, thSCD1 = scene_change)
msadfe= sadfe.Binarize() # binary inverted forward SAD mask

sadbe = MMask(super_even,vbe, ml=ml, gamma=1, kind=1, thSCD1 = scene_change) # backward SAD mask
msadbe= sadbe.Binarize() # binary inverted backward SAD mask

msade = Logic(msadfe,msadbe,"OR") # combined inverted SAD mask
msade = msade.Expand() # expanded inverted SAD mask
msadie = Interleave(msade, msade, msade) # interleaved 3-frame inverted SAD mask

Interleave(cfe,even_fields,cbe) # interleave forward compensated, source, and backward compensated

even_result_fields=DeSpot(show=0,mscene=99,p1percent=3,dilate=3,maxpts=1000,p2=6,mthres=9,p1=12,pwidth=840,pheight=12,mwidth=7,mheight=5,merode=33,interlaced=false,seg=1,sign=0,ranked=true).SelectEvery(3,1)

super_odd = MSuper(odd_fields,pel=2, sharp=1)
vfo = MAnalyse(super_odd,isb=false) # forward vectors 
vbo = MAnalyse(super_odd,isb=true)  # backward vectors 

cfo = MFlow(odd_fields,super_odd,vfo,thSCD1=scene_change) # previous compensated forward
cbo = MFlow(odd_fields,super_odd,vbo,thSCD1=scene_change) # next compensated backward 

sadfo = MMask(super_odd,vfo, ml=100,kind=1,gamma=1, thSCD1 = scene_change)
msadfo= sadfo.Binarize() # binary inverted forward SAD mask

sadbo = MMask(super_odd,vbo, ml=ml, gamma=1, kind=1, thSCD1 = scene_change) # backward SAD mask
msadbo= sadbo.Binarize() # binary inverted backward SAD mask

msado = Logic(msadfo,msadbo,"OR") # combined inverted SAD mask
msado = msado.Expand() # expanded inverted SAD mask
msadio = Interleave(msado, msado, msado) # interleaved 3-frame inverted SAD mask

Interleave(cfo,odd_fields,cbo) # interleave forward compensated, source, and backward compensated

odd_result_fields=DeSpot(show=0,mscene=99,p1percent=3,dilate=3,maxpts=1000,p2=6,mthres=9,p1=12,pwidth=840,pheight=12,mwidth=7,mheight=5,merode=33,interlaced=false,seg=1,sign=0,ranked=true).SelectEvery(3,1)
interleave(even_result_fields,odd_result_fields)
weave()
}

So cool, thank you!

I can see the difference in the photos, it seems a very great advance.
In time: I going for an interlaced video as was the riginal I believe

[Caiosouza]

When I take the sample in VideoRedo, the copy go freezing, as in the first sample

Here sample in mpg:

[johnmeyer]

I don't know what VideoRedo has to do with anything. I also do not understand what you are trying to say.

FWIW, I just tried, and was able to open, in VideoRedo, both your earlier "1996 MX 125.mkv" file as well as the MPEG-2 version, "Grotesque Anomaly 1996 125 MX.mpg." Also, you rendered the MPEG-2 version using the widescreen format, which makes no sense and is the wrong thing to do.

[Caiosouza]

I don't know what VideoRedo has to do with anything. I also do not understand what you are trying to say.

FWIW, I just tried, and was able to open, in VideoRedo, both your earlier "1996 MX 125.mkv" file as well as the MPEG-2 version, "Grotesque Anomaly 1996 125 MX.mpg." Also, you rendered the MPEG-2 version using the widescreen format, which makes no sense and is the wrong thing to do.

What I mean and when I cut and save the segment mkv in the VideoRedo, the sample gets frozen in some frames


In Time: How do I deal with this 720x540 resolution when I'm going to DVD 720x576 without losing the correct aspect ratio 4:3 ?

[johnmeyer]

What I mean and when I cut and save the segment mkv in the VideoRedo, the sample gets frozen in some frames

I think your MKV file has some significant problems. I know this because, since I cannot import MKV directly into my NLE, I first used ffmpeg to convert your MKV video to an HuffYUV AVI video file. However, about 2/3 of the way through the conversion ffmpeg threw an error. I then used the portion of the video that had converted into HuffYUV AVI, and used that. However, even on that truncated video, some of the frames at the beginning and end of the clip were frozen.

So, it is very possible that your MKV file is non-conforming. What this means is that it may have been created with bad software that didn't create an MKV video that follows the proper specifications.

Someone else will need to answer your aspect ratio question. However, when faced with things like this, I just encode to whatever resolution is standard and accept a slight change in aspect ratio. The difference between the two numbers you quoted will be almost impossible to discern, so my recommendation is to not worry about it.

[LMotlow]

In Time: How do I deal with this 720x540 resolution when I'm going to DVD 720x576 without losing the correct aspect ratio 4:3 ?

Didn't you see the scripts I and others posted with two samples earlier? First, you clean up the video (you can forget about cleaning up this disaster, it isn't worth the effort), then resize to 720x576, then encode for a 4:3 display aspect ratio. Is this the first time you've worked with video? Surely you learned something from the previous samples, all of which were the same goofy frame size. They've all been deinterlaced via field decimation, all are over saturated, all have illegal video levels, all are encoded at stingy bitrates, and seem to have come from the same visually disturbed dude.

[LMotlow]

So, it is very possible that your MKV file is non-conforming. What this means is that it may have been created with bad software that didn't create an MKV video that follows the proper specifications.

Not possible, but definitely non-conforming. Nothing in Avisynth or VirtualDub would open it properly, show all frames, or not freeze at some point. I finally got a fully working Lagarith AVI out of TMPGEnc Mastering Works, of all things. Thanks for the revised script.

[jagabo]

Demux the h.264 stream from the MKV. Then open that with AviSynth.

[Caiosouza]

CASE 3 Closed-Is Impossible

[Caiosouza]

FINAL CASE

WellI left the latter for being the most important, I have here a video that was the hardest to get such a rarity, it's not so a bad whole than Grotesque Anomaly , but I need to improve this to put on DVD, the colors seem cold, the brightness ..., levels... maybe, Field lines in a progressive video

I need to make it beautiful, I need to do a miracle here

What filters do you recommend for:

-Improve image quality
-Adjust the colors
-Solve the dirt of field lines
-Maybe a Sharp

[jagabo]

I tried one thing that helped with the high motion shots but looked terrible on the still shots. Basically, it just blurring away a lot of the noise and a giant MCTD so you lose fine detail.

LWLibavVideoSource("1996 MX 125.h264") 
BilinearResize(352,288)
RemoveDirtMC(120)
McTemporalDenoise(settings="very high")
aWarpSharp(depth=5)
nnedi3_rpow2(4, cshift="Spline36Resize", fwidth=720, fheight=576)
aWarpSharp(depth=5)
Sharpen(0.3)

[Caiosouza]

I tried one thing that helped with the high motion shots but looked terrible on the still shots. Basically, it just blurring away a lot of the noise and a giant MCTD so you lose fine detail.

LWLibavVideoSource("1996 MX 125.h264") 
BilinearResize(352,288)
RemoveDirtMC(120)
McTemporalDenoise(settings="very high")
aWarpSharp(depth=5)
nnedi3_rpow2(4, cshift="Spline36Resize", fwidth=720, fheight=576)
aWarpSharp(depth=5)
Sharpen(0.3)

......................Prank?

Any sample?

Maybe I have problems with parts of little movement?

The video is not totally fast moving

[Caiosouza]

"RemoveDirtMC"

I get error, I dont have this, download at?

EDIT: Solved.

About colors, how do I solve this grayish aspect?

[Caiosouza]

Disguising the trash to not lose detail:

DirectShowSource("C:\Users\User\Desktop\Sample 1996 250 World MX.mkv")
ColorYUV(gain_y=-48)
BilinearResize(352,288)
aWarpSharp(depth=5)
nnedi3_rpow2(4, cshift="Spline36Resize", fwidth=720, fheight=576)
aWarpSharp(depth=5)
Sharpen(0.9)
SeparateFields()
Interleave(SelectEven(), SelectOdd().ColorYUV(gain_y=-256))
Weave()
Spline36Resize(width,432).Spline36Resize(width,height)
ColorYUV(gain_y=256)

[jagabo]

Grayish aspect? You mean the first few shots? Most of the video is already over saturated.

[Caiosouza]

Grayish aspect? You mean the first few shots? Most of the video is already over saturated.

Ok, but and how do I solve this?

This?

SeparateFields()
MergeChroma(last, Spline36Resize(width/4, height*4).aWarpSharp(depth=20).Spline36Resize(width,height))
Weave()

[jagabo]

How do you solve what? You want the first few shots to have more saturation? Tweak(sat=x). You should know such obvious things by now.

[Caiosouza]

How do you solve what The video for me looks strain, It doesn't matter, but that's what I see. You want the first few shots to have more saturation? Tweak(sat=x). You should know such obvious things by now.

The video for me looks strain, over saturated? It doesn't matter (It really matters but I do not know). But that's what I see.

I know you're trying to make me think, but I do not know what's wrong, I just know it's ugly, if I could identify exactly what's wrong, maybe I could think of something.
At the same time that some colors seem oversaturated, others look faded, such as the sand for example
Look the colors of Grotesque Anomaly sample These would very close be correct

How do I fix these color levels problems etc.
This whole crap you know.


Do the video colors look good to you? If yes why? If not, why?

[Caiosouza]

It seems more correct? but the red closely to pink...

DirectShowSource("C:\Users\User\Desktop\Sample 1996 250 World MX.mkv")
ColorYUV(cont_y=-35)
Tweak(sat=0.80,StartHue=80, EndHue=115,coring=false)
Tweak(hue=20,coring=false)
Levels(12,1.0,255,16,245,dither=true,coring=false)
ChromaShift(L=-2,C=2)
BilinearResize(352,288)
aWarpSharp(depth=5)
nnedi3_rpow2(4, cshift="Spline36Resize", fwidth=720, fheight=576)
aWarpSharp(depth=5)
Sharpen(0.9)
SeparateFields()
Interleave(SelectEven(), SelectOdd().ColorYUV(gain_y=-256))
Weave()
Spline36Resize(width,432).Spline36Resize(width,height)
ColorYUV(gain_y=256)

[LMotlow]

The colors are horrible. Do you know the difference between green and red? What color should the dirt be? Wouldn't you say that the pics are overly bright, that highlights are badly clipped, and the colors look washed out?

Gamma is far too high, brights extend beyond RGB 255. Partially fixed with Levels() function.
Gain is still too high (brights clipped) and blue contrast too low. Fixed with ColorYUV().
Video has a strong green color cast, some chroma saturation "glow". Fixed with Tweak().
Too much bleeding from over saturation. Partially fixed with FixChromaBleeding() plugin.
Levels analyzed with Avisynth Histogram("Levels"), saturation analyzed with Histogram("Color2").

Original frames on the left, correction on the right:





Change values to suit your personal preference:

AddBorders(0,2,0,2)
Levels(16, 0.75, 255, 16, 235, dither=true, coring=false)
ColorYUV(gain_y=-5, cont_u=100)
Tweak(hue=12, sat=0.95, dither=true, coring=false)
FixChromaBleeding()
## Histogram("Levels")
## Histogram("Color2")

http://avisynth.nl/index.php/Levels
http://avisynth.nl/index.php/ColorYUV
http://avisynth.nl/index.php/Tweak
http://avisynth.nl/index.php/FixChromaBleeding
http://avisynth.nl/index.php/Histogram

More difficult question: What the hell do the makers of these videos think they're looking at? Or they've found ingenious new ways to screw up YUV color. Maybe it's a new fad.

[Caiosouza]

Wow, this was really effective!

You seem to feel very comfortable working with colors, and it works beautifully, just like jagabo works the image, you two should work together!

The only one however was due to the tonality of red that could be closer to the original frame.

More difficult question: What the hell do the makers of these videos think they're looking at? Or they've found ingenious new ways to screw up YUV color. Maybe it's a new fad.

Originally not so to bad I believe, but was spoiled by a number of wrong procedures of capture conversion and copying...

[jagabo]

Tweak(sat=0.50, StartHue=80, EndHue=115,coring=false)

Be careful when using Tweak like that. It creates discontinuities when there are smooth gradients. Before and after:


[Attachment 39667 - Click to enlarge]

[Caiosouza]

Tweak(sat=0.50, StartHue=80, EndHue=115,coring=false)

Be careful when using Tweak like that. It creates discontinuities when there are smooth gradients. Before and after:


[Attachment 39667 - Click to enlarge]

Like this?

[jagabo]

I think so. I'd have to see the source too to say for sure.

[Caiosouza]

I think so. I'd have to see the source too to say for sure.

Source video do not contain this.

This firts tweak not cause This probelm, but kill the red...

Tweak(hue=12, sat=0.95, dither=true, coring=false)

The problem is a minimal alteration....and cause a high "Storm" in the colors....

[LMotlow]

I think so. I'd have to see the source too to say for sure.

Source video do not contain this.

This firts tweak not cause This problem, but kill the red...

Tweak(hue=12, sat=0.95, dither=true, coring=false)

The problem is a minimal alteration....and cause a high "Storm" in the colors....

Not so. The line "Tweak(hue=12, sat=0.95, dither=true, coring=false)" increases red and reduces green. Positive values for Tweak() move the colors toward red. Read up on the link to the Tweak() function and see for yourself.

You'll never correct all of the color in that video. YUV channels were corrupted a long time ago. They'll never be the same again. Perhaps you could try some color blend modes and adjustment layers in AfterEffects or Premiere Pro to experiment with possibilities. Color in that video is ruined, along with a lot of elements. Probably better to let it stay as-is.