Can I fix jitter in post?

Rich Parry wrote on 4/2/2012, 5:20 PM
I shot some video that has some bad jitter (stutter) during pans and tilts. The jitter is not from camera stability, a tripod and fluid head were used. I believe the fault lies with the camera’s inability to keep up with the relatively fast camera movement (aka rolling shutter?). No matter the cause, I am looking for suggestions to two questions.

1) I am looking for ways to reduce the jitter in the video in post.
2) Is there something I can do when shooting to reduce the jitter in the future?

I rendered to two different frames rates. Both suffer from the jitter issue. Video shot with Canon 5D Mark II at 23.976fps.
http://vimeo.com/39663948 (24fps render)
http://vimeo.com/39664272 (30fps render)

Thanks in advance,
Rich

CPU Intel i9-13900K Raptor Lake

Heat Sink Noctua  NH-D15 chromas, Black

MB ASUS ProArt Z790 Creator WiFi

OS Drive Samsung 990 PRO  NVME M.2 SSD 1TB

Data Drive Samsung 870 EVO SATA 4TB

Backup Drive Samsung 870 EVO SATA 4TB

RAM Corsair Vengeance DDR5 64GB

GPU ASUS NVDIA GeForce GTX 1080 Ti

Case Fractal Torrent Black E-ATX

PSU Corsair HX1000i 80 Plus Platinum

OS MicroSoft Windows 11 Pro

Rich in San Diego, CA

Comments

johnmeyer wrote on 4/2/2012, 5:55 PM
Short answer: don't shoot at 23.976 ("24p").

If you shoot in this mode, you will get "judder" when you horizontally pan the camera, unless you pan really slowly. There is nothing you can do about this: it is the nature of a really slow frame rate, and is not a defect in your camera and is not due to rolling shutter.

If you want to make the existing video look better, the only solution is to use Twixtor or other similar motion-estimation software program to interpolate between frames and create an artificial 29.97 ("60i") interlaced format. You can also use this program (or the free MVTools2, via AVISynth) to go to 30 fps progressive or some other faster frame rate. In my experience, until you get to about 50-60 events per second (50i, 60i, 50p, or 60p), you are going to have judder when panning the camera.

The other solution is to not pan the camera except when tracking a moving object, and then do so only when zoomed in quite a bit so as to not make the background (which will still judder, even though the moving object being followed by the pan will not) look too bad.
Rich Parry wrote on 4/4/2012, 8:16 PM
Thank you for taking the time to respond. Your analysis and suggestions are appreciated. I'll look into your solutions to fix my current video, and will shoot at 30fps rather than 24fps, at least until I get a 60fps camera.

Rich

CPU Intel i9-13900K Raptor Lake

Heat Sink Noctua  NH-D15 chromas, Black

MB ASUS ProArt Z790 Creator WiFi

OS Drive Samsung 990 PRO  NVME M.2 SSD 1TB

Data Drive Samsung 870 EVO SATA 4TB

Backup Drive Samsung 870 EVO SATA 4TB

RAM Corsair Vengeance DDR5 64GB

GPU ASUS NVDIA GeForce GTX 1080 Ti

Case Fractal Torrent Black E-ATX

PSU Corsair HX1000i 80 Plus Platinum

OS MicroSoft Windows 11 Pro

Rich in San Diego, CA

johnmeyer wrote on 4/4/2012, 9:16 PM
I'll look into your solutions to fix my current video, and will shoot at 30fps rather than 24fps, at least until I get a 60fps camera.I would recommend 29.97 which is also sometimes called 60i. It is the standard video frame rate in North America, i.e., "normal" video.
farss wrote on 4/5/2012, 3:13 AM
That is particularly bad. I've been shooting 25p lately and some of my pans are around that speed and sure don't look that bad although they are mostly tracking shots.

What shutter speed did you use, anything faster than 1/48th is not going to help judder on pans. The other thing is it looks like you've failed to do a Computer RGB to Studio RGB conversion hence the overcooked look and that also doesn't help judder.

Bob.
johnmeyer wrote on 4/5/2012, 10:14 AM
I agree with Bob that the video you posted on Vimeo looks far worse than it should. I tried downloading the video from Vimeo, but that appears to be disabled for your video. Perhaps if you could upload to DropBox or some similar server, I could take a look at the exact file that you uploaded. I suspect it will show duplicated fields or frames, but I can't really tell without having the actual video.
larry-peter wrote on 4/5/2012, 10:41 AM
johnmeyer may be correct. I once had an issue with a JVC camera where it would intermittently record 24p with duplicated fields (i.e. line-doubling) rather than a true progressive frame and the results looked extremely similar to this.
Rich Parry wrote on 4/5/2012, 10:42 PM
farss,

Thanks for reply. I shot the video at 1/50 shutter speed based on 180 degree rule for shooting 24fps. Do you think I should have shot with slower shutter speed?

You are correct, I failed to do a Computer RGB to Studio RGB. I don't know what that means, I'll have to get out the book and research that.

Regarding the overcooked look, not sure what that means. I used one VP FX, "Sony Levels", maybe the look shows my poor color grading skills and that is what you are seeing as overcooked.

I look forward to anything you wish to suggest. Again, thanks for reponse.

thanks,
Rich

CPU Intel i9-13900K Raptor Lake

Heat Sink Noctua  NH-D15 chromas, Black

MB ASUS ProArt Z790 Creator WiFi

OS Drive Samsung 990 PRO  NVME M.2 SSD 1TB

Data Drive Samsung 870 EVO SATA 4TB

Backup Drive Samsung 870 EVO SATA 4TB

RAM Corsair Vengeance DDR5 64GB

GPU ASUS NVDIA GeForce GTX 1080 Ti

Case Fractal Torrent Black E-ATX

PSU Corsair HX1000i 80 Plus Platinum

OS MicroSoft Windows 11 Pro

Rich in San Diego, CA

farss wrote on 4/5/2012, 10:55 PM
"Do you think I should have shot with slower shutter speed?"

No, you've got the shutter speed so close to correct (technically it should be 1/48th) as to make zero difference. I guess you could try a 270deg shutter to add more blur but I really doubt that'll help. The only time I've used a slower than 180deg shutter is to get a bit more light.

Having got that possibility out of the way then we're down to two possibilities:

1) You simply were moving the camera too quickly for the subject matter. JM's original comments about shooting at low fps are very apt and the people shooting film or digital cinema are always having to be mindful of these limitations.

2) Something else more obscure is going wrong but as JM said we'd need access to the original footage to know for sure.

"Regarding the overcooked look, not sure what that means."

On my office PC monitor it looks too "contrasty" and over saturated. My monitor is set to display video with levels from 16 to 235 correctly, your camera records 0 to 255 and if you graded the video on an uncalibrated / uncorrected computer monitor then what I'm seeing will not be the same as what you're seeing.

Bob.
Rich Parry wrote on 4/6/2012, 12:08 AM
John, Bob,

John you got it. Your "duplicate frame" theory is pretty darn close and just about dead on.

Based on your comment that the video may have duplicate fields, I loaded the "rendered" file on a VP timeline and stepped through it frame by frame. I did not see duplicate frames, but something even weirder.

I found that every other frame was a sharp image that I would expect. However, in between each of these sharp frames was a frame that was very blurry (terribly blurry, horrible, horrible, horrible).

I don't know where the terrible blur is coming from, the blur is not there in the original footage.

The bad news is I have no idea what could be wrong, the good news is I have something tangible and objective to work with. Before my analysis was purely subjective.

I changed Vimeo settings if you wish to download and verify my analysis. I am going to continue my research, but I am thinking something is wrong in the rendering process.

Thanks in advance for your time, any further help is much appreciated. Should that not happen, I got something to work with and will keep you posted if I have some success.

This should make a good weekend project. Thanks.
Rich

CPU Intel i9-13900K Raptor Lake

Heat Sink Noctua  NH-D15 chromas, Black

MB ASUS ProArt Z790 Creator WiFi

OS Drive Samsung 990 PRO  NVME M.2 SSD 1TB

Data Drive Samsung 870 EVO SATA 4TB

Backup Drive Samsung 870 EVO SATA 4TB

RAM Corsair Vengeance DDR5 64GB

GPU ASUS NVDIA GeForce GTX 1080 Ti

Case Fractal Torrent Black E-ATX

PSU Corsair HX1000i 80 Plus Platinum

OS MicroSoft Windows 11 Pro

Rich in San Diego, CA

farss wrote on 4/6/2012, 1:19 AM
Oh dear,
I downloaded the file from Vimeo, the 24fps one.
Both Vegas and MediaInfo report the frame rate as 11.988 fps which is half 23.976.
Me thinks you need to address that problem first.

Also as expected Vegas's waveform monitor shows your levels as being wrong.



Bob.
megabit wrote on 4/6/2012, 1:56 AM
From the Help system on the MC AVC template's "Allow source to control fps" item:

Select this check box if you want the plug-in to optimize the frame rate when encoding.

Does it ring a bell?

Piotr

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

farss wrote on 4/6/2012, 2:18 AM
"Does it ring a bell?"

No but ARGGGHHH :(

Now here's a funny tale of woe.
A few weeks ago I had to devise a waterproof housing for an expensive microwave transmitter to be worn on someone's back whilst on a jetski...in the surf. Tested it once in the workshop no problem. Then it was loaned to a local broadcaster and when it came back for a final check there was this horrible problem with judder that came and went.
After a while we gave up and got the vendor's engineer to come have a look and the people it'd been loaned to had had a fiddle with the H.264 encoder settings, reset them to factory default and we were good to go and yeah, it worked on location too, for two weeks and stayed watertight.
Now it rings a bell :)

Bob.
johnmeyer wrote on 4/6/2012, 9:50 AM
Allow source to control fpsPiotr, I'll bet you are correct.

Those two check boxes should be removed from the encoding dialog. I can't imagine how much video has been ruined over the years by people checking those boxes. I think there are actually some templates included in Vegas that have these boxes checked.

In my experience, the video is always screwed up when these boxes get checked. I suppose they may provide some help with really, really low bitrate encodes, but even then I wonder if they are useful.

My advice: never check either of these boxes.
riredale wrote on 4/6/2012, 10:51 AM
Looking at the 24fps Vimeo version I see a series of sharp stills. The frame rate appears to be very low, like 10 or 12Hz, and I don't see any motion blur.

Based on what others have said here, some setting has caused the very low frame rate. As for motion blur, I suppose that if the full 24fps were shown then the eye would have less time to lock in to individual frames and the presence or lack of blur would not matter.
megabit wrote on 4/6/2012, 11:17 AM
"I think there are actually some templates included in Vegas that have these boxes checked."

John,

these are mainly the templates intended for the Internet formats - like the MC AVC "1080p for the Internet" one I mention. The reason those check boxes are set by default is probably to ensure the lowest file size for Internet uploads; the problem though is that it should not be called "Optimized" but "Compromised" :)

Piotr

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

Rich Parry wrote on 4/6/2012, 11:47 AM
Piotr, John, Bob,

Thanks for all the help. While the new version isn't perfect, it is much improved. I unchecked the "Allow source to control fps" and that resulted in a 23.976 video rather than a 11.988 video.

I rendered with VBR Max=20,000,000 bps and AVG=10,000,000 bps. Does that seem reasonable?

http://vimeo.com/39894929

This new version can be downloaded on Vimeo. Thanks a million!
Rich

CPU Intel i9-13900K Raptor Lake

Heat Sink Noctua  NH-D15 chromas, Black

MB ASUS ProArt Z790 Creator WiFi

OS Drive Samsung 990 PRO  NVME M.2 SSD 1TB

Data Drive Samsung 870 EVO SATA 4TB

Backup Drive Samsung 870 EVO SATA 4TB

RAM Corsair Vengeance DDR5 64GB

GPU ASUS NVDIA GeForce GTX 1080 Ti

Case Fractal Torrent Black E-ATX

PSU Corsair HX1000i 80 Plus Platinum

OS MicroSoft Windows 11 Pro

Rich in San Diego, CA

johnmeyer wrote on 4/6/2012, 12:23 PM
The reason those check boxes are set by default is probably to ensure the lowest file size for Internet uploads; the problem though is that it should not be called "Optimized" but "Compromised" :)I am sure you are correct. However, while having that ability to trade off frame rate for spatial quality may have been a reasonable "engineering trade off" five or six years ago, Internet video is now full frame rate, HD quality, with Netflix, Hulu, Vimeo, and many other sites quite capable of delivering great quality at 24 fps or higher.

Therefore, my recommendation to SCS would be to hide these settings under an "advanced" button, and to re-work all existing templates to un-check these boxes. As this thread shows, people are getting into trouble because of them, and delivering video that looks like heck. This bad quality reflects badly on Vegas and, unfortunately, on the person delivering the video, even though the problem is in no way their fault. How the heck are they supposed to know??

Now that we know that this particular problem was indeed caused by these checkboxes, I begin to wonder how many other similar problems that have been reported over the past few years have also been caused by this? I definitely had the problem many years ago when I first used the AVC templates for Internet encoding and uploading and came to the conclusion that the Vegas encoders were broken. I used external encoders (which still do a much better job) to get around the problem.

I wonder if SCS will ever improve low bit rate encoding to the point where the results are as good as the free Handbrake and MeGUI encoders? It sure seems like a "professional" package should be able to provide better quality video for Internet uploading than what you can get from a free "hack" piece of software.

riredale wrote on 4/6/2012, 4:08 PM
Wow, what a difference 24 versus 12 makes!

As an aside, I believe the vertical cliff is Half-Dome, yes?

I was scared out of my wits one Memorial Day weekend when we camped in Yosemite and hiked to the top of Half Dome one day. Climbing that cable ladder going up the side with nothing, NOTHING, preventing you from slipping and taking a dozen other climbers with you gave me the willies.
Laurence wrote on 4/6/2012, 4:42 PM
It is much better but it still looks pretty jumpy. You didn't shoot with IS (image stabilization) on by any chance did you?
farss wrote on 4/6/2012, 5:17 PM
"While the new version isn't perfect, it is much improved"

Clearly the excessive judder is now gone and what remains is what forms part of the "filmic look" of shooting 24fps.

Having downloaded your footage and looked at it with the video scopes waveform monitor your levels are all over the place. Some shots have the blacks at 0 IRE and others at -10 IRE and the highlights are doing the same, in one shot you've pulled them down to around 80 IRE.

The footage probably doesn't need Colour Correction as such and the traditional CC tools in Vegas I'd suggest aren't the best for tackling such shots and such footage. My goto tool is the Colour Curves. I have to admit they're a pain to work with and have an annoying bug that causes tangents to fly off into never-never land but stick with that tool while keeping an eye on the waveform monitor and you'll find you can do great things in an intuitive fashion.



Note that the "curve" I'm using there is not a correct cRGB to sRGB transformation, it is a traditional "S" curve that's lifting the blacks and rolling off the highlights. Arguably it's achieving little as the highlights (sky) are already clipped. I've used it just to give an example of what can be done with the Color Curves.

Bob.
Rich Parry wrote on 4/8/2012, 12:43 PM
You asked if the vertical cliff is Half-Dome, yes it is Half Dome.

You may be interested to know that starting this year (2012), you need to enter a lottery to hike to the top of Half Dome. I think I heard they are limiting it to 50 hikers/day. Seems like a small number, so I could be wrong. But I know there is definitely a lottery.

Rich

CPU Intel i9-13900K Raptor Lake

Heat Sink Noctua  NH-D15 chromas, Black

MB ASUS ProArt Z790 Creator WiFi

OS Drive Samsung 990 PRO  NVME M.2 SSD 1TB

Data Drive Samsung 870 EVO SATA 4TB

Backup Drive Samsung 870 EVO SATA 4TB

RAM Corsair Vengeance DDR5 64GB

GPU ASUS NVDIA GeForce GTX 1080 Ti

Case Fractal Torrent Black E-ATX

PSU Corsair HX1000i 80 Plus Platinum

OS MicroSoft Windows 11 Pro

Rich in San Diego, CA

Rich Parry wrote on 4/8/2012, 12:48 PM
Laurence,

Regarding you question about IS, if you mean IS in the camera, the Canon 5DM2 does not have IS.

If you mean did I accidently have "lens" IS turned, I don't remember. I shot the video with the Canon 100mm f/2.8 macro lens. I am 99% sure it was off.

thanks,
Rich

CPU Intel i9-13900K Raptor Lake

Heat Sink Noctua  NH-D15 chromas, Black

MB ASUS ProArt Z790 Creator WiFi

OS Drive Samsung 990 PRO  NVME M.2 SSD 1TB

Data Drive Samsung 870 EVO SATA 4TB

Backup Drive Samsung 870 EVO SATA 4TB

RAM Corsair Vengeance DDR5 64GB

GPU ASUS NVDIA GeForce GTX 1080 Ti

Case Fractal Torrent Black E-ATX

PSU Corsair HX1000i 80 Plus Platinum

OS MicroSoft Windows 11 Pro

Rich in San Diego, CA