Image stabilization comparison

riredale wrote on 5/5/2008, 9:58 AM
I came across in another thread the news that a German magazine called c't did a comparison study in the 05/2008 issue. I can't read German and in any event couldn't find the specific issue, but was able to copy a summary table here. Sorry for the crummy image quality, but that's all I was able to find.

As I understand it, the stabilizers were tested under four conditions:

- stabilizing camera turns
- stabilizing continuous shaking/jitter
- stabilizing closeup scenes
- stabilizing single shakes
- stabilizing with default settings.

++ is the best, -- is the worst.

So they liked DeShaker the best, followed by Mercalli. The Premiere Elements stabilizer was poor, and the one in PinnacleStudio was awful.

If anyone can find the article and give us a translation, that would be great. Meanwhile, I'm going to try out that JohnMeyer script one of these days. Using DeShaker manually is certainly doable, but a bit of a pain.

Comments

Coursedesign wrote on 5/5/2008, 10:07 AM
The "ohne Wirkung" rating for Pinnacle means "no effect," i.e. it did nothing whatsoever.

Wonder how these compare with the image stabilization in Avid Media Composer (very easy to use & realtime) and Final Cut Pro (that uses code copied from Shake, he-he)?
jrazz wrote on 5/5/2008, 10:09 AM
Once you use the script, you don't go back :)

j razz
johnmeyer wrote on 5/5/2008, 4:11 PM
I keep meaning to update that script to make it MUCH easier to use.

One of these days ...
JJKizak wrote on 5/5/2008, 4:40 PM
The defaults in the script work fine for me. It made things simple. I select the clip I want to deshake, press the button, and whhaaaaasssssaaaaa! It's done. The two biggest mods are to change the drive for those big files and to change the log drives for Vista 64.
JJK
johnmeyer wrote on 5/5/2008, 5:43 PM
In looking at their comparison chart, one of the things that jumps out at you is how long it takes Deshaker to do Pass 1. I wonder if they used the author's defaults, which take FOREVER. There are all sorts of "cheats" that can dramatically reduce the time it takes to do pass 1. I used a few of these in the defaults for my scripts. If you read the author's documentation (not mine), you'll find out what you can do (change the number of pixels being tracked, and only look at part of the frame, which often also helps track the motion better).

Serena wrote on 5/6/2008, 12:09 AM
DeShaker does give the user a lot of control, more so than Mercalli. But it is slower.
crocdoc wrote on 5/6/2008, 2:04 AM
Complete newbie here, but how do the image stabilisers in Boris FX and After Effects rate?
JeanMovies wrote on 5/7/2008, 8:08 AM
Watch this tutorial to see how to quickly smooth out shaky camera footage with ProDad's Mercalli plug-in.

http://www.studiodaily.com/main/richmedia/9162.html
Soreill wrote on 5/7/2008, 3:31 PM
Wow, that tutorial is almost too good to be true. I like how they compare the workflows with the same footage in other products.
This is exactly what I'd be looking for as time and ease of use are most important to me. If the presets produce results are that good, I don't think I'll need that much more control. Plus it handles different formats with no hassel. I'll be downloading the demo tonight to see for myself.

Forgive my ignorance but can someone explain what "the script" is? Where can you find it?
Coursedesign wrote on 5/7/2008, 4:16 PM
If the presets produce results are that good, I don't think I'll need that much more control.

To quote the comment after the video on StudioDaily.com:

I use Mercalli often with great results. However, this program will cause focus to go soft and is tricky to get the right results. It takes a lot of practice. Biggest problem is in tight shots, as it zooms in to compensate for the borders, so composition is critical. Focus issues are the result of this zoom which can be as much as 135% which means fat pixels.

If you want the best quality, you have to resize manually with the best tools, then stabilize with the best tracker you have available.
johnmeyer wrote on 5/7/2008, 10:09 PM
The other reason for soft focus is the nature of stabilization: when the camera moves, it blurs the motion. You don't notice this in the original video because the moving motion is "expected" to be blurry. Once you stabilize it, however, the video looks like it is going in and out of focus.

One reason I haven't yet switched to Mercali despite its obvious advantages is that Deshaker has a very nice border synthesis algorithm, so you don't need to zoom the video to take care of the inevitable edge problems. Thus, you can have your cake (stable footage) and still be able to eat it (no quality degradation from zooming into the image and thereby reducing total number of pixels in the image).
Serena wrote on 5/7/2008, 10:20 PM
Mercalli is easy to use and quick in execution. DeShaker allows much better designation of the reference areas and smoother correction. It may be that I've set high parameters in DeShaker, which also points to the degree of direct control available to the user. I agree that Mercalli, at least for the examples I've tested, isn't good at border synthesis. But I own Mercalli too (well, not the company, just a copy of the software!) .
crocdoc wrote on 5/8/2008, 2:30 AM
Soreill, after reading this thread I searched and found the script here script

Double click the .exe file and let it unzip to the specified folder, move (or copy) the deshake.js file into your vegas Script Menu folder. Launch Vegas, go to Options/Customize toolbar, find deshake on the left hand side and click 'add' so it becomes a button on your toolbar.

On your Vegas timeline, select the clip(s) you want to deshake and click on the deshake toolbar button (it's just a little script icon, but when your cursor/pointer hovers over it, it will say "deshake"), then watch the magic. When the deshaking is finished, the modified clip will appear as a take so you can hit the T key to toggle between original and stabilised footage, or right click on the clip, choose 'take' and choose between the takes at the bottom of the drop down menu.

By the way, great script,!
Rory Cooper wrote on 5/8/2008, 5:52 AM
The defaults in the script work fine for me

hi JJKizak

what scripts are these? where do i get them?

thanks
Rory
JJKizak wrote on 5/8/2008, 9:02 AM
The script is available just above in "crocdoc's" thread. The script defaults are all you need to get started.
JJK
DGates wrote on 5/8/2008, 3:29 PM
With DeShaker taking longer to process, maybe that's why it's considered better?
riredale wrote on 5/8/2008, 4:49 PM
I doubt it, though I don't have access to the original article. What I like most about DeShaker is the ability to have it "fill in" the black borders with data that, in most cases, is not easily distracting to the viewer. I think this is a better idea than the "smearing" option offered by Mercalli, and the "zoom" option offered by everyone. As mentioned earlier, zooming necessarily makes the entire image softer.

I suspect it's also much more likely that an expert can tune DeShaker to a particularly difficult situation, given all the variables in the configuration menus and the ability to manually go into the log file and make frame-by-frame adjustments, if desired. But the tradeoff is much more complexity, though one can install JohnMeyer's script and have practically no configuration hassles at all.
Serena wrote on 5/8/2008, 9:58 PM
I suspect that DeShaker taking longer (2 passes and 30 frame leader and trailer) is a main reason other mechanisms are supposed superior. I was very taken with Mercalli options and the samples I tested, so I bought it. It works well. Except I found that a couple of my handheld clips that I wanted "as tripod" were rendered with the occasional twitch which was worse than the original gentle waver. Repeating with DeShaker gave me a smooth stablisation. On the other hand Mercalli did better with the transition phases of a handheld pan that was unfortunately rather rough. But the way Mercalli fills borders is very poor and it has to be zoomed in to delete them. DeShaker is quite superior in that regard except when gross changes occur in the image. And DeShaker does have a cost advantage!
Cliff Etzel wrote on 5/9/2008, 6:03 AM
Looking at the options for file output type, I have a couple of questions.

1) Does the script work with HDV content?
2) If so, which VirtualDubCompressor file type option would be the best option when working with m2t files???

Cliff Etzel - Solo Video Journalist
bluprojekt
JJKizak wrote on 5/9/2008, 6:15 AM
The script works well with HDV Cineform Neo files. Don't know about m2t files. Also works in Vista 64 if you modify the six log destinations in the script. You can also modify the script from the "C" drive to another drive to handle huge files as the script defaults to uncompressed.
JJK
Cliff Etzel wrote on 5/9/2008, 7:05 AM
Sounds like uncompressed is the best way to go.

Cliff Etzel - Solo Video Journalist
bluprojekt
Dan Sherman wrote on 5/9/2008, 7:14 AM
"UnShaker" is a hardware solution.
Also known as a tripod.
Properly used it's been working great for 100+ years.
No update required, they pretty much last a lifetime.
DGates wrote on 5/9/2008, 7:28 AM
Ditto that. Keep it as stable as possible during acquisition and you won't have as many problems. None of the software solutions can ever replace that.

I did use SteadyHand from Dynapel a few years back. It did pretty good with some sources of jitter, but not so good at others.
johnmeyer wrote on 5/9/2008, 8:04 AM
Two quick notes:

the way Mercalli fills borders is very poor and it has to be zoomed in to delete them. DeShaker is quite superior in that regard except when gross changes occur in the image.Serena, something I forgot to put into my Deshaker guide and into the script documentation is this VERY important hint that I didn't discover until about six months ago. Here it is:

Deshaker does not alter either the first or last frame of each event that it stabilizes.

Thus the video will always begin and end with the exact, precise X-Y-Z-Rotate location of the original video.

"So what?" you say.

Well, this means that if you find that your initial try at stabilizing your footage doesn't work, you can cut your event into smaller chunks, putting the cut at the point where there is a problem. Then, stabilize each individual chunk. Result? No glitch!! What's more because the last frame of one event and the first frame of the next event are unchanged, the transition from one stabilized event to the next is perfectly smooth. Thus you can stabilize within a scene in "chunks."
"UnShaker" is a hardware solution.

In all these deshaking/stabilization discussions someone always mentions this, as if we are all rank amateurs that don't know what we're doing. Fact is there are dozens of reasons why that solution won't work, can't work, and doesn't work.

1. You have to use amateur footage that someone gives you.

2. The nature of your business requires "run and gun" shooting, for instance shooting a riot -- extreme example, but most news gathering fits into this category (other than the pointless practice of shooting some talking airhead in front of a building where something happened two days ago). Other examples include certain sports where the camera must be in motion (try shooting a sailboat race on a yacht with a tripod, or getting aerial coverage with a tripod). OK, I guess Spot has a helmet-mounted tripod for his sky jumping.

3. You want to do a camera motion shot but don't have the money for steadi-cam. You can actually get pretty darn good results with a DIY steadying device followed by post with some sort of deshaker.

I have lots of other examples. The point is that professional video encompasses a huge variety of applications, many of which do not permit the use of a tripod. Deshaking is therefore one of the most essential tools for many of us, and is one reason why many editing programs now include it as a feature. The idea that its use is somehow the sign of poor technique or a low-class DP just isn't true.