Noise in picture

[sanlyn]

@jagabo, yes, I figured out that missing-parts business. And thanks for that tip on I/B frames, it makes sense.

@pdr, this is the first I 've heard of that requirement for RemoveSpots(). But since I'm still in the world's earliest stages of learning masktools, I might try figuring that one out....well, I'll give it a try, but something like mt_makediff is the only thing I can think of that might work.

No, wait, why should makediff work, as opposed to something else? OK, I give up.

[jagabo]

this is the first I 've heard of that requirement for RemoveSpots().

RemoveSpots() is often problematic with cartoons because of the large motions between frames and stuff that appears only in one frame. There's clear blue sky behind the missing segment in both the preceding and succeeding frames. So RemoveSpots() thought it was a spot. I once had the same issue in a cartoon with a scrolling star field. Most of the stars disappeared after RemoveSpots(). I still have that clip around! Before RemoveSpots() on the left, after on the right:



The stars were moving to the left and down with each frame.

This is much less of a problem with non-animated video since it usually has more detail and doesn't have large motions like that.

[poisondeathray]

@pdr, this is the first I 've heard of that requirement for RemoveSpots(). But since I'm still in the world's earliest stages of learning masktools, I might try figuring that one out....well, I'll give it a try, but something like mt_makediff is the only thing I can think of that might work.

No, wait, why should makediff work, as opposed to something else? OK, I give up.

I think you're misunderstanding me. It's not a "requirement"

What I'm saying is all the dirt/spot removal filters will ALWAYS remove important objects to some extent. Things like eyes, shoelaces, beltholes, fingernails, stray hair, etc.... It depends on your tolerance level whether or not it's "acceptable". Just compare (either interleave(a,b) 2 versions , or compare in different tabs in avspmod). They ALWAYS cause damage to some extent.

The masks I'm talking about are semi-manual in a compositing application like AE, Nuke, etc.. There is no way to do this "automatically" in avisynth "perfectly" . Those filters will either miss some spots, or take away important stuff on some frames. The reason is there is no algorithm that can effectively distinguish between what is a "unwanted spot" vs. an important detail to be retained. The way to do this "perfectly" is to composite layers with masks (e.g. you might use strong filter on one layer, weak on another, medium on another) . It's still way faster than the alternative (e.g. photoshop only)

[sanlyn]

Right, poisondeathray, I've seen the effect with RemoveSpots a number of times and had to use other means in drastic cases. Out of curisoity (again, since I don't know what the heck I'm doing), I tried an experiment with mt_adddiff() in MaskTools2.

In the image below, a section of frame 744. The top image used RemoveSpots(). The bottom image used this code:
rs=mt_adddiff(last,last.RemoveSpots())
In the bottom image it almost looks as if the broken line was restored. But in fact the "break" is filled with a less-black line. And of course the colors are the added difference between the original and the new frame: brights are brighter, darks are darker. Oh, well....I'll understand masktools one of these days.


As it happens, this frame didn't have any unwanted dots.

[brusno]

Now, I have tried to fix the film with some avisynth filters.

Picture from the source on top, and from the filtered film on the bottom:


There are 2 things that I find strange. For the first that the picture from the source is frame 570, while the second picture is frame 572. It's the same clip, the second picture is just from the film loaded with an .avs script.
And for the second, the colour is changed in example nr. 2. Is it Mpeg2Source() which causes the change in colour? The carpet is more red in pic 2. I know it isn't any of the other avisynth plugins, because the change in colour is present even if I remove all the other plugins.

I've used CPU=6 in Mpeg2Source
UnDot()
Deblock(quant=30) and
TTempSmooth()

and I think the result is quite good. The carpet isn't loosing too much details either.

[sanlyn]

Found the culprit: RemoveSpots(). First time I've ever seen that happen.

Wow, you must be joking or sarcastic ??

All the dirt/spot removal filters will do that (remove wanted objects) even with the strength turned way down

They are best used with compositing layers, and masks, never alone

No, I've seen RemoveSpots and even RemoveDirt make some tiny things disappear, but not as obvious as here. Then again, it makes sense: if the filter thought that piece of rope was a no-no, it replaced it with a section of cloud or sky from another frame. Hate it when that happens, 'cause then I have to write code, map out pixels, etc. In the long run, RemoveSpots is a great time saver, but no plugin is perfect. The perfect "Read My Mind" filter or app hasn't been written yet.

[jagabo]

There are 2 things that I find strange. For the first that the picture from the source is frame 570, while the second picture is frame 572. It's the same clip, the second picture is just from the film loaded with an .avs script.

Didn't we discuss this ealier?
https://forum.videohelp.com/threads/353657-Noise-in-picture?p=2222744&viewfull=1#post2222744

And for the second, the colour is changed in example nr. 2. Is it Mpeg2Source() which causes the change in colour? The carpet is more red in pic 2.

This is a rec.601 vs rec.709 issue. If you add ColorMatrix(mode="Rec.709->Rec.601") to your AVS script the colors will be the same. The question is which is right. How did you make the two images?
https://forum.videohelp.com/threads/329866-incorrect-collor-display-in-video-playback?p...=1#post2045830

[brusno]

Ok, then I understand why the frames are different in the two examples.
I took a snap shot in VirtualDubMod to make the pictures.
I'll try the colormatrix in .avs script.

[brusno]

Wow! The ColorMatrix code worked! The colors are similar to source
Is it only necessary when showing the filter-results in mpg2, or must it be used when converting to .mp4 also?

[jagabo]

Wow! The ColorMatrix code worked! The colors are similar to source
Is it only necessary when showing the filter-results in mpg2, or must it be used when converting to .mp4 also?

The colors were probably correct before ColorMatrix. Ie, I suspect the original screen shot you posted had the wrong colors. What did DgIndex say about the colorimetry?

[brusno]

ah...you mean that the colors are being showed wrong in DGIndex when opening a VOB file?
I'm not sure what you are referring to with colorimetri. Nothing showed in DGIndex about that. But under video meny; YUV -> RGB; PC scale.

[jagabo]

ah...you mean that the colors are being showed wrong in DGIndex when opening a VOB file?

I don't know how how you made your PNG images so I can't speculate on what's right and what's wrong. DVD sources are normally rec.601. VirtualDub and VirtualDubMod use rec.601 whenever they converts YUV to RGB.

I'm not sure what you are referring to with colorimetri. Nothing showed in DGIndex about that.

When you save the index or preview the video:


BT.470-2 B,G is the same as rec.601 for our purposes here.

[brusno]

Yes, that is what mine is showing too; BT.470 - 2 B,G*
So that means that the DVD is rec.601

I used "snapshots source frame" (shift-1). In VirtualDubMod.

[brusno]

It is VirtualDubMod which shows the colors wrong here. (Scarf orange).
DGIndex and VLC are both showing what you mean are the right colors (the darker tone).

But when opening the .avs script file in VirtualDubMod, then it shows the right colors. That would be the .avs file without the colormatrix code.

[manono]

Yes, that is what mine is showing too; BT.470 - 2 B,G*
So that means that the DVD is rec.601

The asterisk means the colorimetry isn't specified in the video stream. Notice that jagabo's picture didn't have the asterisk. And if it isn't specified it's almost certainly Rec.601. DVDs are supposed to be Rec.601.

It is VirtualDubMod which shows the colors wrong here. (Scarf orange).
DGIndex and VLC are both showing what you mean are the right colors (the darker tone).

But when opening the .avs script file in VirtualDubMod, then it shows the right colors.

What's that mean? That you first opened the VOB directly in VDubMod and got wrong colors, and then later opened the AVS and got the right colors?

[jagabo]

As far as I know, VirtualDub and VirtualDubMod always convert YUV to RGB using the rec.601 matrix. In AviSynth you can specify which matrix you want with ConvertToRGB(matrix="rec601") or ConvertToRGB(matrix="rec709"). If you don't specify which you want you will get rec.601. If the source specifies the colorimetry you should follow it. If the colorimetry isn't specified and you're not sure which to use the general rule is to use rec.601 for standard definition, rec.709 for high definition.

[brusno]

It is VirtualDubMod which shows the colors wrong here. (Scarf orange).
DGIndex and VLC are both showing what you mean are the right colors (the darker tone).

But when opening the .avs script file in VirtualDubMod, then it shows the right colors.

What's that mean? That you first opened the VOB directly in VDubMod and got wrong colors, and then later opened the AVS and got the right colors?

Yes, the vob and the .m2v file got the wrong color in VDubMod. While opening the vob and the .m2v in DGIndex showed the right color.
So it seems that it isn't the film which have any wrong color. It's just how VDubMod shows it.

When I use ConvertToRGB(matrix="rec601"), I get right colors in VDubMod. This indicates that rec601 is what the DVD is. And when I use ConvertToRGB(matrix="rec709"), I get the wrong colors in VDubMod.
But when I don't use anything it seems that VDubMod is choosing rec709 by default.

But this is not a big problem, just a strange observation.
What is more important, is to get the right filters to get rid of unwanted artifacts in the film.

Also I'm wondering if anyone knows of a program that can show mp4 frame by frame, the same way that VDubMod does with mpg2.

[jagabo]

In VirtualDubMod select Options -> Preferences. Go to the Display tab. Disable the Use DirectX for Display Panes option. Exit and restart VirtualDubMod. I think you'll see the colors correctly now. If this is the case, your graphics card is converting the colors using a rec.709 matrix. Do the same in VirtualDub to avoid future confusion.

[brusno]

@jagabo:
changing the directX settings as you suggested didn't help. Colors are the same.
If the graphics card used rec.709, then the other programs would be affected by this I think.

[jagabo]

If the graphics card used rec.709, then the other programs would be affected by this I think.

No, it depends on where the conversion takes place. A program can get YUV video to the monitor by two methods. It can give YUV to the graphics card and let the graphics card do the conversion to RGB for the monitor. Or the program can convert the YUV video to RGB and give RGB to the graphics card for display on the monitor. That's what the DirectX setting in VirtualDub(mod) controls.

I don't know what version of VirtualDubMod you're using but the version I have (1.5.10.2) always uses a rec.601 matrix to convert YUV to RGB or vice versa. Exactly the same as VirtualDub and DgIndex.

[tesmith47]

hello everyone, i have a similar noise / artifact issue but it is only on video or pictures on line, dvd i play are perfectly good!!!???!!
this is something new, since i changed my modem
comcast said they needed to update system and required a dociss3 modem
[Attachment 17029 - Click to enlarge]
[Attachment 17030 - Click to enlarge]
[Attachment 17031 - Click to enlarge]

[jagabo]

I doubt the modem has anything to do with your problem. Right click on a video while it's playing and select Settings. Change the state of the Enable Hardware Acceleration option.



Exit and restart the browser.

Also, update to the latest WHQL certified driver for your graphics card. And the latest Flash player.

[budwzr]

Whew, ya'll boys been workin' hard.