Image Stabilization: High-tech Help for Shakey Shots

We’ve all sat through video that shook, shimmied and jerked so horribly
it should have been preceded by a "viewer discretion advised"
graphic. In truth, everyone who has ever held a camcorder has shot shaky
video.

To realize the benefits, and avoid the pitfalls of using image stabilization,
we first need to understand how it works. There are two popular methods
of stabilizing video, so we’ll examine each of them separately.



Moving Pictures

Understanding how image stabilization works isn’t very hard. To illustrate,
we’ll consider a story of two corrupt binocular salesmen: Jerry and Larry.
These optical outlaws bilked hundreds of people $145 apiece for fake "super-stabilized"
binoculars.

Jerry and Larry always set up their binocular cart in front of the same
picturesque graveyard. When a victim tried their bogus binoculars, Larry
would jump behind the cart and retrieve a poster-sized picture of the graveyard.
He would then quickly position the picture between the victim and the graveyard.

With sharp eyes and fast hands, Larry would
watch for small movements of the binoculars. If the binoculars moved a degree
to the left, Larry would shift the picture to compensate. If they moved
three degrees up and to the right, Larry would move the picture accordingly.
To the person looking through the binoculars, the image was steady as a
rock.

Though the technology is a little different,
a camcorder’s image stabilization system works in much the same way. If
the camcorder detects unwanted motion, it shifts the recorded image to compensate.
If you move the camcorder a few degrees to the right, the image shifts to
cancel the motion. The result is smoother shots, up to the mechanical or
electrical limits of the system.



Optical and Digital

Camcorders achieve image stabilization though two distinct ways: through
digital processing or optical trickery. The digital approach is the more
simple and cheaper to implement, as it requires no moving parts. Optical
image stabilization adds complexity to the camcorder’s lens, which translates
to a higher cost. This explains why you usually find optical image stabilization
on top-of-the-line camcorders.

Digital image stabilization (often called Electronic Image Stabilization
or EIS) performs its magic using the CCD of the camcorder. First, EIS takes
the central portion of the CCD image–roughly 90 percent of the sensor’s
output–and digitally expands it to fill the whole screen. This frees up
the outer edges of the CCD and allows them to function as a buffer area.
The camcorder can then move the smaller active image area around in any
direction on the face of the sensor, until it reaches the physical border
of CCD.

Imagine if your camcorder had a little joystick that would allow you to
move the active image area around at will. If your camcorder were stationary,
the result of moving the image area would look just like you were panning
and tilting the camcorder. The camcorder itself isn’t moving, but shifting
the active area of the CCD makes it look like it is.

The reverse is also true. If you were to pan the camcorder to the left,
and move the active image area in the opposite direction, the resulting
video would look like the camcorder didn’t move at all. If you could move
the joystick quickly enough to counteract every jiggle of the camcorder,
the resulting video would appear as if the camcorder were stationary. This
is exactly how EIS works: just like Larry and his moving picture with the
binoculars, EIS senses the camcorder’s motion, and moves the CCD’s active
area to counteract the motion. See Figure 1 for a visual explanation of
the EIS system.

Optical image stabilization achieves the same result by completely different
means. Instead of moving a portion of the CCD image in response to camcorder
motion, optical stabilization actually redirects the light coming into the
lens. It does this with movable prism mounted near the front of the lens.

A prism is basically a hunk of glass with splayed surfaces. Light passing
through the prism bends in proportion to the angle between the front side
and the back. If the front and back of the prism are nearly parallel, the
light bends by a small amount. If the angle between front and back is greater,
the light takes a more pronounced turn through the prism.

Camcorders with optical stabilization use an ingenious device called a variable-angle
prism. This prism is nothing more than two pieces of glass separated by
an optically inert fluid, with a flexible bellows all the way around their
perimeter. Small motors collapse or expand the sides of the prism, changing
the angle between the front and back lens. This allows the camcorder to
steer the light moving through the camcorder’s lens. Couple these motors
to an electronic circuit designed to sense a camcorder’s motion, and the
prism can counteract small movements of the image by steering the light
in the opposite direction (see Figure 2).



See It, Feel It

With the means in place to counteract unwanted camcorder motion, all that’s
missing is a method to determine when it’s happening. Camcorders have two
common ways to do this as well, one relying on the CCD and the other on
more complex mechanics. Sound familiar?

The simplest way to detect camcorder motion is to evaluate the output of
the CCD itself. If the camcorder is shooting a subject that’s not filling
the whole viewfinder (which is usually the case), only a small portion of
the screen will be moving at any given time. When the whole image
moves, it’s a pretty good bet that it is the camcorder itself, not the subject,
that is in motion.

When the image stabilization circuit senses the whole image changing in
the same direction, it assumes the camcorder is moving and will attempt
to compensate. Usually, the camcorder will run algorithms that evaluate
how fast and what direction the motion is occurring. These help the camcorder
distinguish between desired movements and unwanted shake.

The other method to detect camcorder motion is to sense it directly. Motion
sensors mounted in the body of the camcorder can detect camcorder motion
on all three axes, and compensate accordingly. As with the optical image
stabilization system, motion sensors are more complex and costly to implement
than the simpler CCD system. Because the motion sensors are actually detecting
the camcorder’s movement, however, this approach is less prone to false
readings.



What if I Want to Pan?

Though they have the almost magical ability to dampen unwanted camcorder
shake, image stabilization schemes do have a dark side. The main drawback
to image stabilization is its tendency to try to eliminate intentional
camcorder motion. No stabilization system can tell the difference between
an accidental bump to the left, and the beginning of a pan. After you’ve
sustained the motion for a second or two, the system can be sure it’s a
desired motion.

This causes many stabilization systems to fight the first few degrees of
a pan or tilt, and continue drifting a few degrees at the end. More than
just unnerving, this can make a mess of your most calculated, professionally
smooth pan. For this reason, many people disable their image stabilization
when intentionally moving the camcorder.

You can also trick the systems that evaluate the CCD to determine if the
camcorder is in motion when a large portion of the screen contains a moving
subject. This could be a train pulling out of the station, a van driving
by, or even just a person moving near the lens. One time, a camcorder tested
by can Videomaker had an interesting reaction when we bounced a large
cardboard box a few feet in front of the lens. With EIS engaged, the box
stayed stationary on the screen while the rest of the world seemed to be
moving up and down.

Finally, certain EIS systems can cause a noticeable drop in image resolution.
This is because they’re creating the image from a smaller block of pixels
on the CCD, and enlarging them to fill the screen. A camcorder with a 280,000-pixel
CCD, for example, may be using fewer than 250,000 pixels to create the image
with EIS engaged.

The best solution is to load the CCD with extra pixels so engaging EIS has
no visible effect on image quality. Several higher-priced camcorders use
this approach, and the results are excellent–it’s truly impossible to detect
a drop in resolution with EIS engaged.



Steady As She Goes

For most people, the drawbacks of image stabilization pale compared to its
many benefits. Playback tells the whole story–image stabilization has made
the world of handheld video a much smoother place to be.

Videomaker
Videomakerhttps://www.videomaker.com
The Videomaker Editors are dedicated to bringing you the information you need to produce and share better video.

Related Content