'Stabilize Media' in VMS10 -- horrbile results..

Sykes wrote on 7/7/2010, 7:23 AM
Has anyone tried the Stabilize Media feature in VMS 10 Suite? I did a 1:00 clip (under General Footage: medium analysis) and the results came out zoomed, slow motion and blurry with lots of noisy artifacts. I've also tried several other settings and various time lines to keep it short and simple, but to no avail.
I'm better off keeping it shaky with higher quality. Is there a way to improve this or am I doing something wrong?

Comments

Birk Binnard wrote on 7/7/2010, 8:27 AM
It's not clear to me either what settings are optimal.

I have a hand-held 360 degree walkaround of the Stonehenge monument in AVCHD format that is rather shaky. I was hoping the stabilizer software could smooth out the video, and it did, but at the same time it made the monument's stones look like they were made out of jello: in the stabilized video the stones appear to wiggle and change shape. But the overall video is less shaky.

This is not the infamous "jello effect" associated wtih CMOS sensors. I was not walking fast enough for this to happen and individual frames do not show the effect.

My guess is my video is just too shaky for the software to fix it properly. After all, there is no such thing as magic, yes?
richard-amirault wrote on 7/7/2010, 9:52 AM
"...I did a 1:00 clip (under General Footage: medium analysis) and the results came out zoomed,..."

I don't have VMS 10, but how do you think anti-shake software works? It *has* to zoom in so that when it adjusts the position on the video to counteract each "shake" the edge of the video does not show on screen.
musicvid10 wrote on 7/7/2010, 10:47 AM
Actually, I think the results are marvelous! Especially for a $94 software.

I guess this has to do with knowing what DIS is actually capable of, rather than what your expectations are.

"and the results came out zoomed, slow motion and blurry with lots of noisy artifacts.

1) Would you prefer black edges and corners jumping all around the frame?
2) The output is the actual framerate. Don't know what you mean by "slow motion."
3) Did you know it has to deinterlace and zoom to do its job? Add a slight touch of sharpening to the output and all will be well.
4) Artifacts are usually the result of bad rendering settings. I am unable to duplicate your observation, using HDV source with lots of fine detail and motion. Worst thing I noticed was a tiny bit of planar skew.

I can't wait until they put basic stabilization in Vegas Pro. What a time saver!

david_f_knight wrote on 7/7/2010, 12:18 PM
I agree with musicvid. The results of the image stabilization software are far better than what I expected -- pretty close to magic. This one new feature alone justifies the upgrade for me. But there are limitations to what it can accomplish, and things the user can do to help it achieve the best results. The version of the image stabilization software integrated within SVMSHDP10 (they really need to come up with simpler names...) is a much simplified version of their (proDAD's Mercalli Expert) plug-in that they sell directly to users. After reviewing proDAD's website, what I really would have liked to do was buy a license to their full-blown image stabilization software and use it with my copy of SVMSP9. Unfortunately, they don't sell a version for Vegas Movie Studio (but they do for Vegas Pro), so the only reasonable option for me was to get SVMSHDP10 and try to live with the significantly reduced functionality.

To help the software give you the best possible results, shoot your video in progressive mode, shoot with a short enough exposure time that there is very little motion blur due to camera shake, but not so short that CMOS rolling shutter distortion due to camera shake becomes a problem (which you may be able to counteract by selecting the rolling shutter compensation option). Experiment, but you might start out with a 1/250 second exposure time. The more stable you hold your camcorder, the better the results.

As a last-ditch manner to improve results (at least to minimize the amount of image zoom the algorithm applies), break your video event into multiple shorter events and stabilize each of them independently rather than as one long event. The events will not likely match up if you have a jump cut from one to the next, so be creative and if possible cut something else in between the clips or use a transition effect. If you have footage with a large but brief jump but then the rest of your clip has a "normal" amount of shake, cutting out the large jump altogether will allow the algorithm to work much better (it won't have to zoom in nearly as much). It's like the weakest link in a chain: the worst shake in a video event dictates how much the algorithm will have to zoom in for the entire clip. So, find a way to minimize the worst shake in each clip before stabilizing it.

Finally, try the different stabilization templates... each sets the parameters a little differently and will produce different results.
Markk655 wrote on 7/7/2010, 5:08 PM
David,

Wholeheartedly agree with you post. Much has been written about Mercalli and what settings are good for it. http://declic-video-fx.com/language/en-US/proDAD/Mercalli.aspx Other discussions can be found at other forums. Your points are spot on.
Sykes wrote on 7/8/2010, 12:47 AM
I did some test runs on cut portions and the results are a bit better, at least the rendering time was a lot less.