Image stabilization - body or lens?

Ciel_Rouge · June 24, 2011

One more important issue before I go to bed here in Europe (6:17 am local time). I am not planning to use a tripod as the nature of the events and my own preference tend to exclude that. What would be better - to have sensor shift in the body itself or to have stabilization in the lens itself? What would give me a better image without the much dreaded blurring?

ziggy53 · June 24, 2011

Both sensor-shift and in-lens stabilization work well in shorter focal length situations. In-lens optical stabilization tends to work better in the higher end (costly) long focal length optics.

Only in-camera sensor-shift technology will help ostensibly with all lenses, albeit better with short focal lengths. Only in-lens optical stabilization will help with viewfinder view and, conditionally, AF acquisition speed.

Neither technology makes up for slow (small aperture) lenses. Neither technology helps with subject motion.

Nikon and Canon do not offer sensor-shift camera bodies.

Larger aperture lenses tend to be a much better general solution and they tend to offer better image quality at all similar apertures, compared to slower aperture lenses of equivalent focal length.

dlacouture · June 24, 2011

And optical stabilization, while supposedly better at longer focal length, can also introduce some strange effects in the out-of-focus parts of the image. After all, it's working by moving one lens element, and thus does change the optical path of the lens.

It's also worth noting that only in-body stabilization can compensate for rotational movement along the lens' axis (but, to the best of my knowledge, only Pentax' K5 and K7 offer this kind of stabilization).

NeilL · June 24, 2011

Ah, Ciel-Rouge, you are strengthening my impression of you as a person of "black and white", "cut and dried" predilections!

D Forgive my presumption!

(And please, get to bed earlier!)

I, and I think I am not alone, do not consider blur bad. On the contrary. If there is one characteristic that is unique to photography it is blur. Painters do not/cannot blur their images. Have you ever seen a sculpture with blur? Or listened to a symphony with blur?

Photography is graced with this most affecting tool, and it is one of the elements in photographs which distinguishes great from the rest.

When you want sharp, without tripod etc, or IS, get the best light possible on your subject, put some distance between yourself and your subject, choose a wider focal length, a higher shutter speed and larger aperture, expose for the right, place the AF spot correctly, choose the most appropriate drive mode, brace yourself, and get into the zone of your much practised best handholding technique.

Like charity, sharp handheld pictures begin "at home".

Sleep tight! (Do you dream in blur?)

Neil

MarkR · June 25, 2011

Neil, I've seen sculptures with blur, seen paintings with blur, and heard symphonies with blur.

NeilL · June 25, 2011

MarkR wrote: »

Neil, I've seen sculptures with blur, seen paintings with blur, and heard symphonies with blur.

I don't doubt it!

Neil

MarkR · June 25, 2011

You all've never seen a sculpture in which features which should be sharp are not well defined? Never looked at an impressionist painting? Heck, Vincent Van Gogh, nuff said! Never heard music in which the notes/melody has been deliberately obscured by other sounds? Things like reverb are an audio equivalent of blur.

Artists blur things all the time, photography is no different.

NeilL · June 25, 2011

richy wrote: »

... a trip to Glastonbury.

or Rochester, NY apparently!

Neil

MarkR · June 25, 2011

NeilL wrote: »

or Rochester, NY apparently!

Neil

Neil, not only have you not addressed my points above, but you're stooping to personal attacks. Back in my college speech and debate days, that was considered bad form and the last resort of a poor argument. Perhaps things have changed since then? :lol

Molotov Everything · June 25, 2011

Mark I'm with you on that. Back in college I had a band that never really went anywhere, but I put a lot of effort in to our songwriting and sound engineering and I immediately thought of reverb, compression and equalization as the audio equivalent of blur. But it kind of works on the inverse, as in you add the reverb, compression and EQ to isolate something whereas in photography you add blur to everything you don't want to isolate. I mean on the one hand, reverb does blur the sound by adding all the extra voices of it, but the practical application, unless you really tweak it out, is to make that sound fuller and therefore more prominent in the mixdown.
Actually really reverb compression and EQ can all be used to emphasize or de-emphasize different things depending on how you set them.

NeilL · June 26, 2011

MarkR wrote: »

Neil, not only have you not addressed my points above, but you're stooping to personal attacks. Back in my college speech and debate days, that was considered bad form and the last resort of a poor argument. Perhaps things have changed since then? :lol

Apologies, apologies, apologies Mark!

bow

Peace, brother!

lust

I only gamin!

But wait!

What you say Mark, I think is correct, at least as far as impressions are concerned.

Consider this, however, that in the type of techniques you describe there will always be an edge/s which can be found on examination. For example, the "blur" of features in a sculpture is a series of edges, the "blurred" notes in music remain nevertheless distinct notes, etc. There are no edges in photographic blur.

There is something unique about photographic blur. To take a sculpture again for example, the more edges that are packed into a unit of space the smoother the texture and the sharper the composite edge. A drop of mercury has a very smooth surface because, if I am correct, it has very fine structure (many small edges, because mercury molecules are very small), and its outline is very sharp to the eye, because of this same fine structure. The perceived edges of materials, inorganic and organic, become sharper, and their textures smoother, the more data there is (organic materials have fractal edges which contain a high amount of data, and when seen at greater and greater distances - in effect increasing the amount of available data - appear sharper and sharper - think of a wellknown image enlarging software product).

However, in the case of light - and actually also in the case of sound - the exact opposite happens. The more data the greater the blur. The edge between light and no light is very sharp, and the amount of data is nill for the dark. The more frequencies of light, ie the more data, at the edge the more blurred the edge. Similarly, the more frequencies of sound at locations of transition the more like white noise, blurred, the sound becomes. Not music, unless white noise does it for you! So for energies, sharpness depends on simplification of data. That is precisely what our senses do to usefully perceive.

To get photography sharp you have to posterise, ie simplify the data, at the transitions. To blur a sculpture you have similarly to simplify the data at the transitions.

Neil

NeilL · June 26, 2011

Molotov Everything wrote: »

Mark I'm with you on that. Back in college I had a band that never really went anywhere, but I put a lot of effort in to our songwriting and sound engineering and I immediately thought of reverb, compression and equalization as the audio equivalent of blur. But it kind of works on the inverse, as in you add the reverb, compression and EQ to isolate something whereas in photography you add blur to everything you don't want to isolate. I mean on the one hand, reverb does blur the sound by adding all the extra voices of it, but the practical application, unless you really tweak it out, is to make that sound fuller and therefore more prominent in the mixdown.
Actually really reverb compression and EQ can all be used to emphasize or de-emphasize different things depending on how you set them.

Yes ME, as I wrote to Mark above, for energies, blurring involves adding more data, just as you describe you did.

Neil

NeilL · June 26, 2011

PS

Having said all the above...

it's possible that the sum of data across a half-black-half-white image is equal to the sum of data across an image with gradations from black to white. A black-white edge has much higher frequencies than a blurred grey edge. Perhaps the higher frequencies contain more data than the summed blurred frequencies, or perhaps they sum to the same?

Perhaps someone who knows their Fourier analysis theory better than me can comment?

Neil

Dan7312 · June 26, 2011

It was not Fourier but Claude Shannon from Bell Labs who showed that the more information you had, the the more bandwidth required to transmit it. In the early days of radio and also in coding theory that was not obvious. In fact at first people thought an audio signal could take up less bandwidth if transmitted via FM than if it was transmitted via AM, which isn't the case.

It's not so much that higher frequencies have more information in them, they could just be noise, but the the more information you want to transmit the wider the bandwith you will need, i.e. higher frequencies such as abrupt transitions between black and white.

So bluring an image does remove information because most (not all though, think unsharp mask) bluring reduces the bandwidth of the image by making transitions less abrupt. Of course selectively bluring an image, bokah, makes the information left in it more dear, which is probably 'togs like it:D

Fourier is about representing a signal as the sum of a set of sine waves and the Nyquist theorm says what the highest frequency sine wave in that set will be under certain conditions, so that is sort of like what Shannon was talking about. But it was Shannon who did the foundational stuff in information theory and the bandwidth need to transmit it. A picture is just a way of transmitting a information.

Well, back to our regularly scheduled programming

NeilL wrote: »

PS

Having said all the above...

it's possible that the sum of data across a half-black-half-white image is equal to the sum of data across an image with gradations from black to white. A black-white edge has much higher frequencies than a blurred grey edge. Perhaps the higher frequencies contain more data than the summed blurred frequencies, or perhaps they sum to the same?

Perhaps someone who knows their Fourier analysis theory better than me can comment?

Neil

NeilL · June 27, 2011

Dan7312 wrote: »

It was not Fourier but Claude Shannon from Bell Labs who showed that the more information you had, the the more bandwidth required to transmit it. In the early days of radio and also in coding theory that was not obvious. In fact at first people thought an audio signal could take up less bandwidth if transmitted via FM than if it was transmitted via AM, which isn't the case.

It's not so much that higher frequencies have more information in them, they could just be noise, but the the more information you want to transmit the wider the bandwith you will need, i.e. higher frequencies such as abrupt transitions between black and white.

So bluring an image does remove information because most (not all though, think unsharp mask) bluring reduces the bandwidth of the image by making transitions less abrupt. Of course selectively bluring an image, bokah, makes the information left in it more dear, which is probably 'togs like it:D

Fourier is about representing a signal as the sum of a set of sine waves and the Nyquist theorm says what the highest frequency sine wave in that set will be under certain conditions, so that is sort of like what Shannon was talking about. But it was Shannon who did the foundational stuff in information theory and the bandwidth need to transmit it. A picture is just a way of transmitting a information.

Well, back to our regularly scheduled programming

Thanks for this post Dan, really interesting!

I wonder, and suspect, if the transmission of a representation of things, which you describe, is equivalent in terms of the data/energy values involved, to the things represented. I can understand that to increase transmission signal to noise ratio would necessitate greater bandwidth. But, to take a sound example, for an orchestra to produce an approximation of white noise all instruments would be playing at once, and at as many frequencies as possible. But to produce an "edge", one note on the piano would suffice. A lot of data and energy, versus a little.

I think my thinking is a little whacky, but it's late and I haven't had dinner yet! Why go back to the regular program when this diversion is so fascinating?

Neil

Dan7312 · June 27, 2011

Some people might say that all you have to do to get an orchestra to produce white noise is give them a score by Arnold Schoenberg

, not me but some people...

A signal that goes from zero to some level, any level different than zero actually, in zero time contains an infinite amount of energy. Of course in real life you can't do that.

Piano notes are really complicated. The hammer strikes the string at a well chosen point and what it produces is partly governed by Ohm's law, not the one you might be thinking of. Ohm's law says there will be no overtones present in the note, that have nodes at the point where the string is struck. But then the hammer bounces back and and suddenly it is a completely different string that is trying to vibrate. And then if you have John Cage sticking pencils in between the strings it gets even more complicated.

I think a piano note is the most complicated musical note in terms of Fourier representation as sine waves, except for the human voice.

Making white noise is hard, I don't think an orchestra could really do it. White noise means that the Fourier representation there is a sine wave for every possible frequency, an infinite number of them, all at the same level.

Hermann von Helmholtz in "On the Sensation of Tones" did the fundamental work in figuring out why a basoon doesn't sound like a piano even when the play the same note. Even though it was written over 150 years ago I think Dover still has it in publication. It is a scientific paper but not that hard to read.

In terms of energy vs. information I'm pretty sure that more info means more energy, because it means more bandwidth, and high frequencies mean quciker level transition which requires more energy.

Hermann von Helmholtz also did work on how we perceive color, so I guess this post can be considered to be vaguely related to photography:D

NeilL wrote: »

Thanks for this post Dan, really interesting!

I wonder, and suspect, if the transmission of a representation of things, which you describe, is equivalent in terms of the data/energy values involved, to the things represented. I can understand that to increase transmission signal to noise ratio would necessitate greater bandwidth. But, to take a sound example, for an orchestra to produce an approximation of white noise all instruments would be playing at once, and at as many frequencies as possible. But to produce an "edge", one note on the piano would suffice. A lot of data and energy, versus a little.

I think my thinking is a little whacky, but it's late and I haven't had dinner yet! Why go back to the regular program when this diversion is so fascinating?

Neil

Image stabilization - body or lens?

Comments