Stabilizing Your Video Footage, Part 2

Table 1. Full ratings for all contenders.

Stabilizing shaky footage and smoothing out camera motion are problems that you'll periodically face. Last time out, I looked at how well Apple Final Cut Pro's SmoothCam filter and Apple Motion's Stabilize function performed. In this article, I'll look at Red Giant Software Magic Bullet Steady ($199) and the Boris Continuum Optical Stabilizer Mac ($99). I was going to include the Image Stabilizer in CHV-Electronics MotionTracker-collection ($69), but the workflow was very complex and the tool performed very poorly in smoothing tests (though adequate at stabilization), so it wasn't worth your time or mine.

Table 1 summarizes all findings for both reviews, including the comparative analysis and render time for my 27-second Apple ProRes test clip. With all tools, I considered two objectives: one stabilizing a clip to make it look like it was shot on a tripod, the other smoothing my pans and tilts to make them look as if shot with a Steadicam rig.

Figure 1. Red Giant Software Magic Bullet Steady.

Let's start with Red Giant's Magic Bullet Steady, which costs $199. With this tool, you choose your analysis, stabilization, and quality options, then Analyze the video footage, and then render in Final Cut Pro as normal. Analysis can be automatic or manual, where you choose a region in the video that isn't supposed to move. For example, I tested stabilization with concert footage that I wanted to appear as if it was shot with a tripod. In manual analysis mode, I identified a microphone as an object that was supposed to remain stable.

Related Links

Stabilizing Your Video Footage, Part 1 In this issue, I'll discuss the image stabilization task in general, and analyze the usability, quality, and performance of Final Cut's SmoothCam filter and Motion's Stabilize behavior. In essence, these are the free tools that you likely already have...

There are multiple stabilization options: You can detail how much horizontal, vertical, and rotational motion to remove, with the following options for each type of motion removed.

• Fast Shake corrects only shake. It removes high-frequency motion variation and random shake. This option gives the most correction.
• Wobble corrects shake and wobble. It removes medium-frequency motion variation. This is the default setting for both controls.
• Weave corrects shake, wobble, and weave. It removes low-frequency motion variation. This option offers the least correction.
• None will not apply any motion smoothing. There is no correction beyond the correction that happens in the Analyze pass and Automatic/Manual setting.
• All Motion (locked) removes all motion. The resulting footage is completely locked and stabilized, as if the camera had been on a tripod. It may be tempting to use this setting for both axes since it's the strongest correction. But if you fully correct the motion, depending upon your footage, you may completely lose framing of the footage because the footage must be scaled so much to correct for all motion.

There are two quality options, Linear and Cubic, with the latter being the slower but higher quality option. There's also an Auto Zoom option that cuts off any edges created by the stabilization effect, a nice option that I used for all tests.

Let's start with stabilization results. As stated, my goal with the band footage was to make it look as if shot with a tripod. There wasn't a huge amount of motion in my original footage, so I started with Weave for horizontal, vertical, and rotational correction, and the stabilization was minimal. Then I tried All Motion (locked) for all three correction options, at the highest quality setting. Magic Bullet removed most—but not all—of the camera motion, but also introduced some jerkiness in the corrected footage. The bottom line was that Apple Motion was much easier to apply and worked better in these tests.

My testing efforts got me thinking about Magic Bullet's speed advantage, which is significant over Apple Motion. Balanced against this, however, is the experimental time necessitated by the multiple options offered in the Magic Bullet program. In addition, if you switch to another program while analyzing—say Command-Tab-ing over to Mozilla Firefox to check email—Magic Bullet stops rendering, which really stinks with longer clips. Though Magic Bullet is the fastest overall program, the necessary experimentation, and the need to stay in Final Cut Pro during the analysis phase reduce its speed rating to Good.

For smoothing, I used two test clips: one a slow zoom into the Lincoln Memorial in Washington, D.C., the other a pan and zoom around a Porsche exhibit at the Munich Pinakothek der Moderne modern art museum. In both of these jobs, SmoothCam proved quite functional, and was reasonably speedy as well, though if you're stabilizing a small portion of a larger clip, you may have to render it as a QuickTime Reference Movie to avoid analyzing the entire clip. Long story short, potential workflow issues notwithstanding, SmoothCam set a high bar for functionality with these two clips.

As it turned out, SmoothCam's bar proved too high for Magic Bullet to jump, primarily because while both tools corrected horizontal, vertical, and rotational motion, SmoothCam did a better job with these corrections and also seemed to correct speed—in particular, smoothing out the motion of the Munich museum clip to an amazing degree. A quick check of the Magic Bullet documentation revealed this blurb: "The plug-in will not correct camera motion that is caused by an uneven zoom in or out." This seems to indicate that the Magic Bullet effect isn't designed to smooth camera motion temporally, which SmoothCam either does or seems to. Assuming that you want this capability, SmoothCam is a better option.

Either SmoothCam or Motion should deliver equivalent or better quality than Magic Bullet in stabilization and smoothing tasks, and Magic Bullet's speed advantage will likely be at least partially offset by increased tinker time with the controls. It's worth a look if you can't get the results you want from SmoothCam or Motion, but I'd definitely start with the Apple tools and try Boris before Magic Bullet.

Figure 2. Choosing whether to stabilize or smooth in Boris Continuum.

Boris Continuum Optical Stabilizer ($99) is competent, and it offers features unavailable in the Apple products, but its complex interface makes you wish that it came with an Easy button.

Related Links

Stabilization Routines Compared View side-by-side comparisons of SmoothCam, Boris, and Magic Bullet.

Here's the workflow. After applying the effect, you choose your Mode, with two basic functions: Stabilization and Smoothing. The former is designed to lock down your footage, the latter to smooth motion. Three other visual modes let you see the motion vectors in the video, see the motion samples in the video, or manually set up a region within the clip to analyze. This latter feature lets you exclude regions that don't provide motion cues that well, like the sky or water, which speeds your analysis and makes it more accurate.

Figure 3. Choosing what to track during stabilization and edge handling options.

Next, you choose whether to to track X and Y axis motion only (Translation), or track X and Y plus Rotation, or those three plus Zoom. If you decide not to zoom into the video to eliminate any black bars caused by the smoothing, you can choose a color to fill the gap, leave it transparent, repeat pixels on the edge, or reflect or tile the pixels adjacent to the edge. I almost always zoom, but if you prefer not to, these can be useful options that aren't available from either Apple alternative.

Figure 4. Choosing Auto Scale and scaling quality options.

Then you choose whether to Auto Scale or not via a checkbox, and then your rendering quality options. I used the highest possible setting for all tests. Boris also exposes some Optical Flow configuration options which I didn't experiment with, but might prove useful if you're working with problem footage that isn't resolved using the default settings. Also available is flicker reduction, which I didn't need, and a Mixing option which I kept at the default setting.

Nothing too onerous complexity-wise, but count on spending a few minutes reviewing the helpful PDF file that comes with the program before stabilizing your first clip.

Figure 5. In my stabilization trials, Boris zoomed in more than Motion by a significant margin.

In stabilization trials, Boris worked well, but with Auto Scale selected, the effect zoomed into the clip more than necessary to eliminate the black edges, as you can see in the comparison with Motion in Figure 5. Not a huge deal, but scaling always involves some quality loss, so you want to scale as little as possible. I was able to fix the problem by unchecking Auto Scale and adjusting the positioning and scaling manually, but auto is always better, and Boris should have been able to match Motion and my manual efforts automatically.

In smoothing comparisons, Boris does a nice job eliminating horizontal, vertical, and rotational motion in the video, but it doesn't smooth out the speed of the pans and zooms, which SmoothCam appears to. This kept Boris from getting an Excellent in this category.

I also ran into some stability issues with Boris. Specifically, when creating my comparison sequences (like that shown in Figure 5), if I added the titles and overlays before applying the effect, it crashed during the analysis or render stage. As a simple workaround, I could apply the effect and analyze, and then apply the titles and overlays, which worked fine.

Overall, while Boris wasn't up to Apple's quality in my test clips, the program offers several features unavailable in either Apple product, like configurable edge appearance, the ability to adjust optical flow controls and a user definable analysis range. You can try the program before you buy for several weeks, so if SmoothCam or Motion isn't working for you, or if you need a faster solution, give Boris a try.