Curing Camera Shake: Your Guide To Image Stabilization

If your footage sometimes looks as if you taped it during an earthquake, it’s not your fault. No hand-held video will be perfectly steady; even shots made on a tripod can get the jitters when you’re shooting at full telephoto. But you can banish bad vibes from your programs by using image stabilization.

Image stabilizer systems can greatly reduce the shakiness of your images; but to use one you have to invest in a camcorder that includes one. So before you consider doing that, you’ll want to know how well image stabilization works, whether it’s worth the expense and what your options are in selecting a system.

In this article, we’ll look at the kinds and causes of camera shake, the two very different methods of compensating for it and another way entirely to control vibration. Let’s start with the whys of camera shake.

All About Camera Shake

Video images shake because the camcorder that shot them was moving; that’s obvious enough. But different kinds of movement cause different species of twitch.

The most common cause of movement is simply the human muscular system. We hold our arms still through tension between opposing sets of muscles, flexors and extensors. This results in a continuous tug of war–the effect is a slight but constant movement back and forth.

To demonstrate this, point one finger at a word on this page, then close one eye and study the finger with the other. You’ll see that it’s plain impossible to keep the finger from moving slightly with respect to the word. The same thing happens when you hold a camcorder.

The second biggest cause of camera shake is movement from place to place. Every time you take a step you send a small shock wave through your body that causes what you see to bounce slightly. You don’t notice this because your brain compensates for the bumping image, but your brainless camcorder dutifully records a picture that jumps with every jarring step.

The last major source of image shake is vibration, a resonance that causes repeated bumps at rapid and regular intervals. The most common source of vibration is machinery–a car engine, for instance.

In some circumstances, you may suffer two or three kinds of shake at once. When you’re hand-holding in a moving vehicle, for instance, you have to cope simultaneously with contending muscles, bumps in the road and engine vibration.

It’s important to understand the various kinds of movement. Why? Because to work well, image stabilizing systems use different methods to damp different types of shake. As we’ll see shortly, their controlling microchips are more effective at detecting and compensating for certain frequencies and amplitudes of motion. In other words, the odds of unwanted motion finding its way into your video depends on how often it happens and how strong it is.

The most common shake frequencies lie between one and 20 per second. Typically, a slight hand-and-arm tremor will fall between three and five cycles per second (c.p.s.). An out-of-tune car engine loafing along at 1,200 r.p.m. will vibrate at 20 c.p.s. (assuming one vibration cycle per engine revolution).

Practical Consequences

The type of camcorder shake affects the type of image jitter that results. Slow, slight arm movement creates a mild wavering that is not always unpleasant. A very rapid, regular vibration will make the entire image appear to go out of focus. In between, a bumpy walk or ride can create an irregular lurching movement that may actually make some viewers queasy.

The effect on the image is also dependent on the focal length at which you set the camcorder lens. At extreme wide angle, most camera shake is acceptable or even unnoticeable. But at full telephoto, you magnify the camera movement right along with the subject. If you’ve ever tried to hand-hold while covering a football or baseball game from a distance, you know that the results may look as if you were taping from a rowboat in heavy seas.

One other factor affects camera shake: the weight and design of the camcorder. All other things being equal, heavier full-size VHS camcorders tend to be more stable than their VHS-C and 8mm cousins. The reason: simple inertia. The more weight, the more force it takes to move it.

In the design department, camcorders that permit users to hold them away from their faces are easier to keep stable. Your hands and arms can act as shock absorbers, but the bony rim around the eye you press against the viewfinder transfers every vibration from your body to the camera. On the other hand, when you press a full-size camcorder to your cheek during shooting, the solid trunk of its body increases stability. This is one reason most ENG (electronic news gathering) camera operators still prefer the full-sized camera units.

So now that we’ve surveyed the types and traits of camera shake, we can look at how different manufacturers try to compensate for it. There are two basic approaches: electronic and optical.

Electronic Image Stabilization

As its name implies, electronic image stabilization (EIS) reduces shake by manipulating the image electronically. Here’s how EIS operates.

In a camcorder equipped with electronic image stabilization, the control circuitry does not record the full image striking the camera’s light-sensing chip. Instead, it records about 90 percent of the chip’s area. When the camera is still, that 90 percent is centered left to right and top to bottom, as you can see in Figure 1a. [[[Figure 1]]]

Now suppose that the camera jerks to the left. That makes the image shift to the right side of the recorded frame and the result is a shaky picture (Figure 1b).

But if the camera has EIS, then things change. When you bump the camera, the used portion of the image readout area on the chip itself electrically shifts opposite the direction of motion to compensate for the amount of movement, thereby following the subject of the shot (Figure 1c).

How does the camera know when and where to shift the recorded portion of the chip image? By analyzing one of two different kinds of information: 1) changes in the picture itself; or 2) changes in the position of the camcorder. (Each particular camcorder uses only one of these two approaches.)

Image analysis. One method for detecting camera motion is to use a microchip to analyze changes from one picture field to the next. (As you know, one video frame consists of two interlaced fields.) The camera’s circuitry includes two “field memories,” each of which briefly stores one field. That is, one memory stores field one, then field three, then five. The other stores two, four and then six–and so on.

Basically, the microchip looks for movement by comparing selected areas of the second field in each frame with the same areas of the first field. Here’s what it looks for:

  • If the image in some areas differs from field a to field b, but the image in other areas does not, then it’s the subject matter that’s moving, not the camera. For example, the subject walks across the frame but the background stays still.
  • If the image in all areas differs by the same amount, that means the entire image is shifting–and that may mean camera shake.

So if the comparison of the two fields does not show uniform change, the EIS circuitry does nothing. If it does, the chip analyzes the direction of the movement and shifts the active segment of the CCD in the opposite direction. When the image zigs, it zags, by precisely the same amount.

Note what we said earlier: that a uniform change across the whole field may mean camera shake. Sometimes large amounts of movement can be caused by, you guessed it, the movement of a large subject. This explains the greatest disadvantage of most EIS systems–they can’t tell the difference between camera motion and the motion of a subject that fills most of the frame.

Motion sensing. Where some EIS systems analyze image change from field to field, others sense and interpret movement of the camera itself. Tiny, sensitive motion detectors report each physical shift of the camcorder. Because they’re not analyzing the image at all, motions sensors don’t get fooled by subject movement.

But they can still get fooled. What if the camera is moving on purpose–panning or tilting or tracking? What keeps any EIS scheme from busting its hump trying to compensate for large camcorder movements?

Warm and Fuzzy Computer Chips

“Fuzzy logic,” that’s what. Fuzzy logic is a special type of instruction set that allows a computer to guesstimate. Normally a computer must have an absolutely perfect “if/then/otherwise” situation in order to function. For a much-simplified example, imagine a car controlled by computer. Through its programming the computer knows that if the light’s red, then stop the car; if the light’s otherwise, keep going. At the next intersection, the computer talks to a sensor on the surface, “Is the light red or not?”

“Not,” replies the sensor, and the car continues.

But at the next crossing the computer asks, “Red or not?” The sensor reports: “There’s no stop light here.”

“That’s not red,” says the computer, and it pilots the car right into an accident.

But with fuzzy logic programming, the computer might reply instead, “No light? Okay, then is anything approaching on the cross street?”


“Then it’s probably okay to go, so I’ll risk it.”

Fuzzy logic, in short, enables a computer faced with uncertain circumstances to go with its best shot–which in a camcorder is a steady shot. If the entire image continues to change uniformly, the chip realizes that this is camera movement rather than shake, and it doesn’t try to compensate.

Fuzzy logic is also useful in detecting and compensating for different types of shake. Here we return to frequency and amplitude. By analyzing and identifying the particular type of shake, the microchip can optimize its compensation.

Early attempts at image stabilization were limited by the fact that they were effective only within a narrow range of frequencies. A system optimized for arm movements at four or five cycles per second could not respond fast enough to rapid vibrations. On the other hand, a system designed to damp vibrations tended to overreact to slower movements.

In today’s stabilization systems, however, the control circuitry is able to fine-tune the response to match the type of shake.

Though we’ve been discussing motion sensors, control chips and fuzzy logic in the context of electronic image stabilization, exactly the same things apply to optical stabilization as well. But if the sensing and programming are essentially the same, optical compensation is a totally different approach.

Optical Image Stabilization

Optical image stabilization works before the image hits the CCD, so no electronic adjustments are necessary and the image can fill 100 percent of the chip surface.

In fact, optical systems do their work before the image even enters the camcorder’s lens. In principle, the idea is simple: position an optical prism between the scene and the lens to bend the shifted image back on center.

Simple? Suuure! A prism refracts light along just one single axis; but image shake varies infinitely in both direction and amplitude. No single prism could possibly compensate for all types of quiver and quaver.

So the engineers at Canon devised an ingenious solution: a “soft” prism whose axis of refraction changes when you bend it. To create this variable prism, they position two pieces of optical glass with a gap between them, enclosed by an accordion-pleated barrel (see Figure 2).

[[[Figure 2 here]]]

They fill the space between the glass elements with a liquid silicon that has a very high index of refraction.

Expanding the accordion tube at any point around its rim compresses it on the opposite side, which changes the angle between the glass elements. If you think of it as a clock face, expanding the rim at nine o’clock tilts the glass toward three o’clock; expanding at ten o’clock tilts the glass toward four o’clock and so on. Expand the rim slightly for a narrow angle; expand it more fully for a wider angle.

The result: a prism whose axis is infinitely variable over 360 degrees and whose angle of refraction is also variable. In practical terms, this means that whichever way an image deflects, the prism can direct it back at the center of the CCD.

To see how this works, look at Figure 3.

[[[figure 3 here]]]

Figure 3a shows the image of a man and woman directed by a lens onto a CCD. (Remember that camcorder lenses are vastly more complex than this simple diagram.) In Figure 3b, the lens deflect downward slightly, so the image has moved up in the frame and changed position on the CCD. The result appears as image shake.

Figure 3c shows what happens when you interpose the Canon variable prism. By expanding at the top just enough to compensate for the image shift, the prism refracts the image back to the center of the CCD. As a result, the subjects don’t shift in relation to the frame, and the image shake is effectively canceled.

A Short Side Trip

Designed to compensate for camera shake, both electronic and optical systems attempt to undo its effects after they happen. Another approach to stabilization: keep camera shake from happening in the first place. Instead of stabilizing the image, this method seeks to stabilize the camcorder itself.

The most well-known camera stabilizer is the Steadicam by Cinema Products, an ingenious use of spring-arm system that dampens camera movement. Full-size Steadicam harnesses for motion picture work are so big that the operators wear them. And though the results can be spectacular–floating with an actor through a door, up three flights of stairs, down a narrow corridor and through another doorway–they require considerable strength and dexterity to operate, along with plenty of practice.

The Steadicam compensates for massive camera shake such as from running up stairs. It is in a completely different league from camera image stabilizers designed to compensate for shaky hands.

However, in recent years the Steadicam JR has become a popular method of stabilization. Designed for cameras weighing under four pounds, the Steadicam JR operates by floating the weight of the camera on an exoskeleton gimble arm held by the operator. This isolates the camera from the operator’s movements.

Though Steadicams are actually camera rather than image stabilizers, we’ve included them briefly here for the perspective they offer on the pros and cons of electronic and optical methods.

Pros and Cons

All three stabilization methods offer a mixture of positive and negative characteristics.

Electronic stabilization systems are compact because they add no bulk to the lens, and fast because they don’t have to physically move anything. All the heavy lifting occurs electronically, at ultra high speed.

On the down side, many electronic systems sacrifice image quality because they use only 90 percent of the CCD. As a result, the image must be electronically magnified to fill the remaining 10 percent of the frame, with unavoidable loss of sharpness.

To address this problem, manufacturers are now moving to oversize chips, on which 90 percent of the area is equal to 100 percent of a conventional chip. The JVC model GR-SZ7, for example, delivers lossless EIS thanks to a CCD that contains 570,000 sensors!

Another problem with some units is that their effectiveness varies considerably, depending on the frequencies being damped.

Optical stabilization systems don’t require expensive oversize CCDs because they utilize the full chip area. Models we’ve seen, like the Canon ES1000 Hi8 camcorder, boast remarkably consistent results over a wide range of vibration frequencies. The ES1000 effectively damps vibrations from 1 to 20-plus cycles per second; between three and 15 c.p.s., its compensation is over 90 percent effective.

To be sure, optical systems do pay a small penalty in weight and bulk. And, theoretically, at least, they can’t respond as rapidly as electronic systems because they have to move the physical components of their prisms. But these drawbacks are all but unnoticeable in actual camcorder use.

As for the Steadicam JR, the effects you can achieve with it are truly remarkable–as if your camcorder were riding the back of a swooping bird. On the negative side, it’s expensive as accessories go (around $400, street price). And even in its smallest form, it adds considerable size and bulk.

Also, some users report that the unit has a disconcerting mind of its own when you try to move it around. But perhaps it would be fairer to say that even the most junior Steadicam takes a fair amount of practice to master.

In general, all three systems add more or less weight, cost and complexity to your video outfit. But the good news is that they work great! Compare similar footage shot both with and without image stabilization and you’ll see the obvious difference. (For an in-depth report on field-testing stabilized camcorders, see Robert J. Kerr’s Image Stabilizers in the August, 1993 Videomaker.)

Stabilizer Genealogies

The next question, then, is who makes what? Where do you go for an electronically stabilized camera–or for one with an optical system? Here’s an informal rundown on some manufacturers and their product offerings.

Electronic image stabilizing systems are available from Hitachi, JVC, Panasonic and Mitsubishi.

Canon developed the optical system; its third new and improved incarnation appears in models like the ES1000 mentioned earlier.

As for Sony, well, you have to check out each camera model, because this manufacturer sells three different systems:

  • More modestly priced units employ EIS with the classic electronic field image analysis system.
  • Upscale a ways, you’ll find models like the Hi8 TR400. It still uses EIS, but with camera motion sensors instead of field analysis.
  • Up there in white tie and tails is the Sony Hi8 TR700, which uses true optical stabilization licensed from Canon.

And as if that weren’t confusing enough: Sony’s optical system is reported to be the first Canon generation, while Canon itself is up to version three, as noted above.

Who Needs It?

When all is said and done, do you really need image stabilization?

The short answer: if you do a lot of sports or wildlife videomaking–or anything else, for that matter, where you park your lens at full telephoto and keep it there–go for stabilization. The improvement in image steadiness will be dramatic.

If your hands are just naturally somewhat unsteady (and face it, most of us were not born to be neurosurgeons), you will probably benefit consistently from image stabilization at all lens focal lengths.

Otherwise, check it out. At your friendly neighborhood video store, hook a stabilized camcorder up to a big monitor, set the zoom lens to full telephoto and play with the system. Then switch off the image stabilization and check out the difference on the monitor.

If you decide you want an electronic field analysis scheme, verify that the camcorder you choose has an oversize CCD so that you don’t sacrifice resolution.

It would be ironic, wouldn’t it, if you had to lose image quality in one department in order to gain it in another?

Videomaker contributing editor Jim Stinson makes industrial videos, teaches professional video production and writes mystery fiction.

The Videomaker Editors are dedicated to bringing you the information you need to produce and share better video.

Related Content