Geek Articles

How Image Stabilization Works In Camera and In Lens

Published January 30, 2019

Image Stabilization comes in a variety of different names and types. Whether it’s called O.I.S. (Optical Image Stabilization), VC (Vibration Compensation), VR (Vibration Reduction), IBIS (In-Body Image Stabilization) or just IS (Image Stabilization), it all foundationally does the same thing – controls the effects of camera shake to produce sharper images. With the recent years, In-Body Image Stabilization has been created, and the bulk majority of the latest lens releases from Canon and Nikon come with some iteration of image stabilization. But what does all this mean, and how does image stabilization fundamentally work?

Why You Might Need Image Stabilization

Within your first year in photography, you’re likely going to learn a foundation rule of photography; while handholding, to avoid blurry images from camera shake, your shutter speed shouldn’t be slower than your focal length. So if you’re shooting with a 50mm lens, you’ll want to shoot at least 1/50th of a second to avoid camera shake. 200mm lenses should be shot at 1/200th of a second or higher, 400mm lenses at 1/400th and so on.

However, this rule changes entirely once you add Image Stabilization systems into the mix. Most modern IS systems offer 3-5 stops of image stabilization, meaning where you once were theoretically limited to 1/200th of a second on a 200mm focal length lens, you can now shoot the same images at 1/13th of a second (4-stops of exposure). This has enormous advantages, especially when working handheld or with limited available light, which is why every camera and lens developer is working to extend image stabilization to 6 stops and beyond.

Because cameras are a three-dimensional tool, image stabilization systems need to work on up to six different planes to properly correct camera movement. The most simple camera shake will be directional shake; horizontal, vertical and forward/back shakes. Rotational shake, or commonly referred to as pitch and yaw, control the horizontal and vertical rotational movements that can occur while handholding.

 

How Lens Image Stabilization Works

Lensrentals.com, 2018

By default, image stabilization comes in two different flavors – in lens stabilization or in-camera body stabilization. These two platforms work differently, but work to produce similar results. To put it simply, in lens stabilization has a floating lens element which is controlled electronically by a microcomputer and shifts in the opposite direction of the camera shake, helping to stabilize the image. All of this is detected in mere microseconds and can give you up to 5-stops of stabilization, depending on the lens, movement, and focal length. Below is a short diagram showing you how this works to help counteract any camera shake.

In-lens image stabilization is by far the most common type of stabilization system. However, there is another type of image stabilization system that is becoming more and more popular, commonly called In-Body Image Stabilization (IBIS).

How In-Camera Stabilization Works

In recent years, through the help of Sony and Fuji cameras, In-Body Image Stabilization has become more and more common in cameras. Whereas image stabilization within the lens has a floating lens element helping to counteract the camera’s movement and shake, In-Body Image Stabilization has a floating sensor that helps neutralize any movement within the camera. The key advantage to this system is that if your camera has IBIS, all of the lenses you use with it will also have image stabilization.

Is In-Lens or In-Body Stabilization Better?

A common call we get here at Lensrentals.com is the cut to the chase “Which is better?”. But it’s not as simple as that, as both systems have advantages and disadvantages. For example, in-lens stabilization will generally perform better on longer focal lengths, because camera shake requires more compensation at the pivot point (camera) than it does within the lens. This is why many Sony telephotos still have in-lens stabilization, despite having IBIS on all of their mirrorless systems. So let’s look through some advantages of each system, to determine what works best for you.

Advantages to In-Lens Stabilization

  • It’s far more effective in telephoto lenses. A subtle shake of the camera is pretty drastic when shooting at 500mm, and will naturally be better compensated within the lens rather than the camera body.
  • Lens stabilization works better in low light conditions. Because the IS is working as an independent unit, you’ll have better results with in-lens stabilization while in low light conditions. In-Body Image Stabilization will often have trouble metering and focusing in lower light situations while activated.
  • By in large, In-Lens Stabilization is more effective. While many camera companies developing IBIS will deny this, generally in-lens stabilization will provide better results. This is because the image stabilization is fine-tuned for each lens, and usually offers multiple IS modes depending on the situation. However, with systems like the Sony a7rIII and Sony a7III offering 5 stops of image stabilization, this argument is slowly fading away. 
  • Has no effect on your metering and autofocus. Unlike with IBIS, in-lens IS will have no negative effects to your autofocusing and metering while activated. 
  • By design, In-Lens Stabilization will offer better battery life. In-lens stabilization requires smaller motors to move the optics for camera shake, and is far less draining on the battery when compared to in-body image stabilization.

Advantages to In-Body Stabilization

  • Generally, In-Body Image Stabilization (IBIS) is cheaper in the long run. While IBIS will usually be an added cost to the camera body purchase, it is a one time purchase and will usually result in lower lens prices, when compared to similar lenses with IS built in. 
  • In-Body Stabilization is universal – and works with all lenses. To further the point above, once you have IBIS, you should be able to use image stabilization with all the lenses in your kit.
  • Unlike with most lenses with IS built in, IBIS operates in silence. If you’ve activated image stabilization on a lens, you’ve likely heard clicking and other noises from the lens while focusing. That is (usually at least), the image stabilization system making adjustments. 
  • IBIS offers cleaner bokeh when engaged. With IS turned on for in-lens systems, you’re asking the lens to make optical adjustments to counteract any movement, which can result in some weird bokeh. Because the optics are stationary with IBIS system, you will get a cleaner bokeh.

Misconceptions Regarding Image Stabilization

There are a few misconceptions with image stabilization systems that we need to answer often through tech support calls. So let’s go over a few of them here. 

Can you use both in-lens stabilization and IBIS?

In short, yes. While it is dependent on the camera system you’re using (for example, Panasonic has a list of compatible lenses), but you should be able to use them together. With Sony systems, activating both systems will delegate 3-axis stabilization to the IBIS, and leave the pitch/yaw adjustments for the Optical Steady Shot (O.S.S) in-lens stabilization. Fuji systems, at least the Fujifilm X-H1, work in a similar fashion; delegating specific axis’ to different systems to achieve the standard 5-axis stabilization.

Should I turn off IS before demounting a lens?

As a general practice, yes. If you have Image Stabilization activated on a lens, you’ll want to turn it off, wait three seconds, and then unmount the lens. Not doing this can potentially put the IS system in what we call an ‘unparked’ position, which means the optics are still floating, which could cause damage if shaken and jarred. 

Is there a theoretical limit to image stabilization?

Olympus seems to think the limit is 6.5 stops of image stabilization. In a recent interview, Setsuya Kataoka, part of the Imaging Product Development Division at Olympus, claimed that image stabilization’s theoretical limit is set at 6.5 stops of stabilization, due to rotation of the earth interfering with gyro sensors. I’ll let the comments below determine if that is a scientific fact, or just marketing mumbo jumbo.

Does image stabilization help with fast moving subjects?

No. Image stabilization is designed to control only the movements from camera shake. It won’t help stabilize any blur caused by moving subjects.

The Naming Schemes of Various Image Stabilization Systems

Likely because of patents, each brand has their own naming for their image stabilization, which is why most modern camera lenses have half a dozen letters slapped to the end of their official product name. So here is a quick reference guide for what each major brand calls their image stabilization system.

Lens Brand Image Stabilization Name
Canon IS (Image Stabilization)
Nikon VR (Vibration Reduction)
Sony O.S.S. (Optical Steady Shot)
Panasonic Mega O.I.S. (Mega Optical Image Stabilization)
Power O.I.S. (Power Optical Image Stabilization)
Dual I.S. (Dual Image Stabilization)
Sigma OS (Optical Stabilizer)
Tamron VC (Vibration Compensation)
FujiFilm OIS (Optical Image Stabilization)
Olympus IS (Image Stabilization)

 

Hopefully, we were able to help with any questions you may have had regarding image stabilization, and if you have additional questions, feel free to chime in in the comments below or give us a call.

Author: Zach Sutton

I’m Zach and I’m the editor and a frequent writer here at Lensrentals.com. I’m also an editorial and portrait photographer in Los Angeles, CA, and offer educational workshops on photography and lighting all over North America.

Posted in Geek Articles
  • The camera should be turned off before unmounting an IS-equipped lens. I got my 70-200 II damaged by unmouting “hot”. Cost about $80 to repair via CPS.

  • Michael Clark

    “While IBIS will usually be an added cost to the camera body purchase, it is a one time purchase and will usually result in lower lens prices, when compared to similar lenses with IS built in.”

    So where are all of these cheaper lenses?

    If one looks at prices from third party lens makers for versions that fit systems with IS and versions that fit systems with only IBIS, there’s usually very little to no price difference.

    If one looks at comparable lenses in terms of focal length, maximum aperture, amounts of aberration correction, etc., very often the non-IS lenses in the IBIS system are more expensive than the comparable IS/VR lenses from Canon and Nikon.

  • zogzog

    Is the advice to turn off a lens’s stabilizer before dismounting it relevant if you make it a habit to turn off the camera first? The camera should park the IS unit as a part of the shutdown process.

    I always turn off the camera when switching lenses because I was taught that changing lenses while the camera is switched on is a bad idea, regardless of the stabilization issue, because an electrically active sensor attracts dust more easily. (Well, OK, I forget sometimes, but I feel really bad when I do so.)

  • Ricardo Rei

    Hi Ilya,

    The chart would be a spectrum of mechanical waves of maximum frequency of 30Hz. So the power would be related indeed to the displacement (amplitude of the wave, so amplitude of movement) x force (that induces that displacement), that’s energy not power. To get power you should divide by the time interval in wich the force is aplied. Using dB you divide a power P by some reference power Po (10 log(P/Po) so having a fixed time interval of force aplied in the test (equal in P and Po), power in dB will be related to displacement x force. Perhaps…
    The article is complex to me, so just pick some hints to contribute regarding the shake frequency.

    But there are standard methods by CIPA using representative frequencies of 0,1-0,5-1-5-10 Hz:
    http://www.cipa.jp/image-stabilization/documents_e/VMVA_for_DC-X011_E.pdf#page=8

    More on that google:
    “image Stabilization Performance” site:http://www.cipa.jp/ filetype:pdf

  • Ilya Zakharevich

    I presume you mean 2.5.2.

    The graph does not mark what is “Power” (is it related to the maximal displacement, or to the maximal speed?). Makes it not very practically useful…

    Well, since at 0 the density is not 0, then probably it is “the displacement”. However, again, there are dB, and dB: did he take “square of the displacement” when calculating “power”?

    The formula which follows the graph is pure BS. (It grows quadratically with t for a short time t of exposure. As if you press the shutter release only near the extremum of the displacement…)

    Did not try to read any more…

  • Lester

    image stabilization systems need to work on up to six different planes Ah, no. Strictly, a 3D coordinate system has three planes and three axes. IS works in five movement components: pitch, yaw, roll, left/right, and up/down. Not forward/backward.

    “Rotational shake, or commonly referred to as pitch and yaw, control the horizontal and vertical rotational movements that can occur while handholding.” Sadly, there is no sense in which pitch or yaw “control” any movement.

    “Below is a short diagram showing you how this works” I’m not sure what a “short” diagram is. Perhaps “simple”?

    “In recent years, through the help of Sony and Fuji cameras, In-Body Image Stabilization has become more and more common” Sadly, completely misleading. Minolta, Pentax, and Olympus were there long before Sony or Fuji came along.

    “This is why many Sony telephotos still have in-lens stabilization” Not just Sony!

    “with systems like the Sony a7rIII and Sony a7III offering 5 stops of image stabilization” Not just Sony!

    “Can you use both in-lens stabilization and IBIS? […]Panasonic […] Sony […] Fuji” Well, as I understand it, Olympus currently holds the world record for best stabilization, so might be worth mentioning here.

    Zach, this piece comes across as a Sony fanboy post as well as being technically flawed. Not what I expect from Lensrentals. I’d suggest you seek the advice of an independent editor when posting “factual, informative” articles. Apart from that, keep writing!

  • Ricardo Rei

    Hi Roger!

    Hi to all!

    Ok here goes to all science nerds…the article “Image sharpness” from mr. “Falk Lumo”:
    http://www.falklumo.com/lumolabs/articles/sharpness/ImageSharpness.pdf

    For me the article is complex. The text goes trough factors that play some role in image sharpness. One of the factors listed is shake (point 2.5) on page 26 a chart is presented depicting the power distribution of free-hand shake for a typical human photographer (the source is mentioned but i couldn’t find if it results come from some empirical test). So a spectrum from 0-30 Hz is expected, but higher frequency disturbances are likely to be less harmfull than the lower ones that are powerfull. For the article purpose the peak between 4-10 Hz is considered, a formula relating image blur to shake frequency is presented, so for simulation purpose a dominant frequency should be adopted or simulation must be run for each relevant harmonic. So in each axis, shake would be sensed as a spectrum of frequencies, but the system should filter for powerfull harmonic, and apply the right action in real time, in order to reduce the amplitude of the dominant frequency.

    Another interesting question mentioned in point 2.5.1 is introducing the distance to subject into equantion. Translational shake (I think he refers to X or Y axis movement) can be neglected if the distance to subject >100 focal length (so if you have a 200 mm lens, translational shake can be neglected if the distance to subject is higher than 20 m). In this condition: “The same camera displacement then leads to only ?1% blur if translational when compared to rotational, e.g., a 5 px blur measure has less than 0.1 px translational error”.

    As far as I know, when combining OIS+IBIS some manufacturers let Yaw+Pitch for the OIS and Roll+Y+X to IBIS. So I assume that Yaw and Pitch shake would be better corrected with OIS. OIS would correct 2/3 of the worst shake problems that arise when you shoot a large telephoto and your subject is far (the other 1/3 is roll and only IBIS could deal with it).

    Kind regards

  • Good point Ilya. I was approaching it from the physiological point of view, the frequency of human tremors. 3 and 6 HZ are basal ganglion and cerebellar frequencies, there are different likely frequencies in flexar (say biceps) and extensor (triceps) muscles, etc. Whether any of them have the predominant effect, I have no clue.

  • hywelphillips

    In practical terms I have found that using IS with a telephoto lens on a tripod is still well worth doing if there’s much wind, especially if the wind is gusting. It seems to induce some vibration in the system which IS is good at taking out. My experience is primarily with an A7RII and 70-200 f/4 Sony lens – it noticeably improved the critical sharpness of the shots turning both body and lens IS on. In a calm studio on a heavy tripod, though, it really SHOULDN’T be necessary to turn IS on, and I know some photographers claim it degrades the images.

    Personally I leave my IS on all the time, since I usually handhold indoors and shoot out in the mountains outdoors. I can’t say I’ve ever noticed a degradation which I’d attribute to IS. The only thing I definitely switch it off for is astrophotography.

  • David Bateman

    Nope I think you come off as being extremely ignorant. Because of Olympus pushing hard with IBIS since the E510, others have noticed.
    Fuji has one camera and has not introduced IBIS in any newer models.

  • Ilya Zakharevich

    Without further evidence, I do not believe much in these 3Hz–6Hz numbers. KM’s response curve had efficiency decreasing approximately linearly below 8Hz (where it was about 8x), and the IBIS was still working quite well (about 2 stops, IIRC). Then, I interpreted this as the bulk of the shake being above 8Hz…

    On the other hand, with relatively short exposures, 3Hz would give effect 10x smaller than 30Hz (given the same amplitude). So this depends strongly on the target range of exposure times too!

  • Ilya Zakharevich

    There is an ISO testing procedure, but the last time I checked (5 years ago?) it was beyound a paywall. I think they were trying to emulate a hand-tremor…

  • Ilya Zakharevich

    Won’t help even if all the objects in a scene are completely fixed w.r.t. each other: you NEED 1sec exposure for a photo, how much would you need to detect a motion which is so slow?!

  • Wild Wild


    Likely because of patients

    Don’t ya mean *patents*?

  • Daniel Morgan

    Completely agree here Neven, I was about to write something similar.

    Definitely kind of missed the mark on the development of in-body shake reduction.
    You could at the very least update the article with the Pentax brand and what they call their IBIS system!
    .. and I’ve never had my Pentax system have a problem metering or focusing while stabilization has been activated… I’m curious what brand that happens with? If that happens with Sony or one of the other major brands… all the more reason to perhaps mention Pentax in your listing!

    The Pentax system can even be used for long exposures doing astrophotography, for pixel-shifting and taking multiple photos to get better color rendition when used on a tripod, for anti-alias filter simulation as-needed (thus eliminating the need for a physical AA filter/ can produce better detail than systems with an AA filter. ) Lots of things to consider.

  • DP

    >>> Case in point

    there is no case in point because reasons not to do multi-shot in 2004 when first IBIS camera appeared were not about moving sensor by a sensel, but rather because there was no reason to push for it… there are a lot of different IBIS implementations

  • My only answer is I don’t know. I did some research (looking up papers, not lab) just trying to get a feel for what type of motion a human holding a camera would have. It was very complex, and depended immensely both in frequency and direction on position of the camera (held up to the eye is really different than looking at the LCD for example); age of the photographer, presence of numerous minor disease states, medications and drugs, et.

    The one thing I recall is they were almost all pretty low frequency, though, with several peaks around 3 and 6 per second, and I don’t think any really reached 30 / second from the human neuromuscular side of things. Of course, modes of transportation and stuff would be a lot different.

  • Franck Mée

    Once again, that’s EXACTLY what I wrote. I don’t get why you answered me like I was wrong to say that IBIS can provide other functions.

  • Les

    Yes, you can have both features on the same body, but neither one requires (or “provides”) the other. The reason for that is multi-shot moves the sensor by one pixel (micrometers), while IBIS moves the sensor much more, so the implementation is very different. Case in point, multishot was around for nearly a decade before being offered on a camera that happened to also offer IBIS. And the opposite is also true. Correlation does not imply causation.

  • “The laws of physics do not know whether you are making a photo to publish in astrophotography magazine, or in Playboy.” This has now become my favorite quote ever.

  • DP

    actually when Sony entered with the first dSLM + IBIS body Olympus already had more sales of dSLM + IBIS cameras… so while Sony contributed it was a gradual process, even now Sony does not sell a _magnitude_ more IBIS bodies than the m43 … what Sony did is it forced Nikon to finally accept that dSLM + IBIS are the real game… and Canon will be forced to follow now

  • DP

    > And the camera has no clue that the room rotates about this!

    it actually does as it can try to analyze the image readout from the sensor (may be not during the exposure time with current state of consumer sensors for dSLMs or for dSLRs with mirror up, but at least during the framing – using main sensor in dSLM or metering sensor in dSLR)… not that it solves the practical issue of stabilizing during the exposure, but in principle for sufficiently long exposure (close to a second) you can have a lot of readouts… once (if) sensor tech allows multiple sensels readouts during the actual exposure time then you can try to use that to acquire additional point of reference for stabilization.

    PS: or information acquired during framing can be used to adjust IBIS operation during exposure

  • DP

    except a camera can readout 30…240+ and more times image off the said sensor (or part of the frame) and can try (AI is a popular word) to assume that something within that image as being immobile and then try to adjust IBIS operation somewhat… gives additional reference…

  • Ilya Zakharevich

    Actually, you caught me up on this?—?for a few minutes I thought that this may work! What happens: Foucault’s pendulum-in-a-camera is not possible!

    There are two principal components of Foucault’s pendulum: the intertially-stabilized platform (the pendulum) and the markings on the floor. The markings show us that the floor is rotating w.r.t. the intertially-stabilized platform.

    A camera with IBIS already has an intertially-stabilized platform?—?and it is much better stabilized than pendulum (for example, the stabilization does not depend on latitude!). However, the camera has no clue about “markings on the floor”: it cannot find out that the room is rotating about it!

  • Ilya Zakharevich

    As I said above, KM did.

    This single plot captures all the relevant performance information …

    Wrong. Contemporary systems are non-linear.

  • Ilya Zakharevich

    When in school, the log-scale exposure meter I designed for my darkroom would compensate the inverse current I? of a diode (to change I?(exp(V)-1) to I?exp(V)) up to 6–7eV. This was at room temperature (with steady-state heating system?—?not PCM-heating) on the timescale of about 30min–1hour.

    So the achieved compensation depends on application (no surprise!). Remember that there is no feedback loop in image-stabilization, everything is done with forward-control (sp?) only. Having a precise enough model of the dynamical system of IBIS+sensor is really crucial.

    BTW, I know that Konika-Minolta published the spectral curve of their IBIS (the level of compensation as a function of vibration frequency); it was around 10x in the range about 8–80Hz. I also know that Sony uses a non-linear system (to avoid bumps at the limit displacements). I have not seen any other numeric data for more contemporary stuff!

  • Ilya Zakharevich

    The laws of physics do not know whether you are making a photo to publish in astrophotography magazine, or in Playboy. What IBIS knows about is the movement of camera in the coordinate system immobile w.r.t. the fixed stars. It can do nothing else since inertial navigation knows nothing about other coordinate systems

    For either journal, with 100mm focal length, if the camera is stabilized w.r.t. the fixed stars, the image of ground objects would move about 7.5µm/sec on the sensor.

  • Ilya Zakharevich

    There is nothing “ficticious” about Coriolis force (unless you believe that Newtonian mechanics is the final answer) if you use certain coordinate systems?—?but nobody forces you to use these coordinate systems.?—?Moreover, working in these coordinate systems is very error-prone (which you just have shown above!)

    Look at a camera in the coordinate system which does not rotate w.r.t. fixed stars. There is no Coriolis force. The room about the camera rotates; gyroscopic IBIS ensures that the camera behaves as if it does not rotate. And the camera has no clue that the room rotates about this!

  • Steve Tschopp

    GPS and a compass would give you correction for earths rotation. When I do electronic reduction for some source of systematic error 90% is fairly easy to achieve. This is between 3-4 stops. 6.5 stops is 99% correction for the error. At least in the stuff I build I never get to 99% error reduction, might be based on an optimistic limit set by experience. I would be curious to know how many “stops” of noise reduction the latest Bose headphones have.

  • DP

    > IBIS can be associated with multi-shot, but the two aren’t necessarily linked.

    better wording: multi-shot can always be implemented with IBIS, but it does not mean it will be…

Follow on Feedly