Archive for The Motor Forum "We are mature men in the highest cadres of our careers"
 


       The Motor Forum Forum Index -> Photography
..

Interesting...quantum dots and exposing to the right.

I read a couple of things recently that I thought photographers here might be interested in...

The first is a new technology which might appear soon, quantum dots...

http://www.eetimes.com/news/desig...ml?articleID=224000253&pgno=1

http://www.technologyreview.com/communications/24840/?a=f

The second is something that I was aware of but this explains it better than I could...

http://www.luminous-landscape.com/tutorials/expose-right.shtml

Basically, sensors don't seem to be at their optimum if the histogram image is central or shifted to the left and there seem to be advantages in exposing so that the histogram image is shifted to the right. I've been experimenting with this for a while and I recently tried some +1 exposures at ISO 3200. Once the images had been corrected (-1) in RAW there was a clear noise advantage over images shot with a smack in the middle (correct) exposure. The improvement was really noticeable and although I do think that a properly exposed full image (uncropped) ISO 3200 shot from my camera is useable exposing +1 to the right and then correcting -1 post capture made things a lot better, noise really was reduced.

Anyone shooting at higher ISO's should give this a try if they haven't already, the only thing to watch is that there is a possibility of blowing highlights when shooting +1 but this might not be something to get too worried about unless a large area blows out.
..

This was taken at ISO 3200 +1 exposure and corrected -1 in RAW. The second shot is a 100% crop. There is noise visible in the crop but as I haven't done any noise reduction I think that this is a very good result and that it compares well for noise with some professional 7D shots on the net at lower ISO.

I'll definitely use this +1 and then correct -1 in RAW technique in the future.

Dunno if I'm talking to myself here...but just in case I'm not - anyone shooting higher ISO's should give this technique a try if they haven't already. There's a bit about it in this weeks Amateur Photographer mag.



Boxer6

The crop definitely makes the image Alan - IMVHO of course! I can see what you mean about the noise, I think, but its not enough to detract from the overall image.

To my shame, I don't really play about too much with the ISO settings on my camera, but I'm doing a course later in the summer which should (hopefully) educate me somewhat!
..

You will almost certainly know that ISO is "film sensitivity."

The higher the ISO number the more light sensitive your "film" (or sensor) is. A higher ISO enables you to take a picture faster and thus helps you to keep your shutter speed reasonably high and avoid camera shake, particularly if taking a shot in low light.

The disadvantage of higher ISO is that as the sensitivity of your film or sensor increases it's dynamic range falls. You get a greater dynamic range at lower ISO and a much reduced dynamic range at higher ISO's so overall, it's much better to use lower ISO settings.

I hope that you enjoy the course.
Clunes

One thing that is important to note is that ISO is not directly related to noise in the simplistic way most seem to believe.

Noise is caused by lack of light reaching the sensor  and 'exposing to the right' essentially increases the light reaching the sensor (reducing noise)...BUT it also (generally) reduces shutter speed so you could probably shoot at 1600 (rather than 3200) at the same shutter speed and get the same result.

But...what i said above is too simplistic so you can probably ignore that as well  

Seriously though - the shot you took Alan is a bit unrepresentative of a scenario in which you would choose to use very high ISO - there is a good amount of light in that scene so a high ISO shot will look good as there is lots of light still reaching the sensor.

I'd guess you didn't need 3200 in order to get the adequate shutter speed to avoid blur/shake.

If you shoot a scene with lots of very dark areas you will se noise in those areas no matter what!

To round off...noise you can see at 100% on a computer screen is irrelevant in the real world if you print your pictures. 'Noisy' pictures on a screen can look perfectly good printed - even at large sizes and if you are printing small then it is almost irrelevant until the noise completely destroys detail etc
..

All very valid points Clunes.

The shot is just a test shot as I didn't want to post a people shot as not everyone is happy to have their picture on the net. I know this from experience and that it results in my ears aching and sometimes a sudden pain in my upper arm or shin.

There is a lot of light in the shot but it's coming from the window and the room side of the little dragon thingy (a prezie from Vietnam) was in shadow and a higher ISO or more probably a lower shutter speed would still have been needed. I could have used a lower ISO here but the point was to generate high ISO noise and then demonstrate the results. Exposing to the right and then correcting in post processing seems to push the noise out, you can see it reduce as you move the slider to -1, and I'm pretty sure that it's worthwhile especially if using the next highest ISO can be avoided and you can avoid blowing highlights. I'm sure that I couldn't have got a shot this clean at ISO 3200 without exposing to the right, but you are right that sometimes it just doesn't matter. See the example taken at ISO 3200 with a spot on exposure, no noise reduction, but if you zoom in the noise is higher than in the exposed to the right shot.



I'm not too sure about noise being caused by a lack of light reaching the sensor as if that was the case ISO 100 shots would be noisy in the shadow areas and although they are to an extent all things are relative and I think that noise is definitely higher in higher ISO shots as what you are really doing is increasing the gain and it's associated noise and as the number of photons is relatively fewer this leads to the signal to noise ratio being relatively higher and to noise appearing not only in the shadow but even in the midtone and highlight areas.
Clunes

You've kinda contradicted yourself :)

ISO is merely increasing the gain (it has nothing to do with 'sensitivity' in the classic sense of old when film was used)

You said it yourself...the number of photons (light) is fewer which leads to the SNR reducing (low light = less signal) - leading to noise...so it's the lack of light that is the cause of noise...the ISO is not causing the noise.

ETTR is totally valid and a great technique in your example as you are keeping ISO fixed BUT by ETTR you are increasing the light reaching the sensor...so you are increasing the signal and can recover the 'correct' exposure in pp (as long as you don't blow the highlights)

You can shoot very noisy shots at ISO 200 and very clean shots at 1600 - in low light chroma noise can make an ISO 200 shot look terrible 1600 would be less 'noisy'

Ultimately you want to maximise the light reaching the sensor - ETTR is one way to do this and is valid but one needs to be careful and not blame high ISO for high noise. Its the fact that high ISO is used in low light (the real cause) that ensures it gets the blame...

and, of course, there are various other things in play such as the algorithms in the camera etc that are not linear in their operation...
Matt

I don't know if it was luck or the sensor of the 20D being low on pixels for its sensor size, but I managed to recover blown highlights from this, which amazed me.




Off-topic - I don't use the 20D much at the moment as I'm still loving the G10.

I recommend this flash for anyone who doesn't need a large one. 55 from Amazon. It's a Nissin Di28.

him

^^^ Does the Di28 work with the Canon G6 too?
Boxer6

Is that the same Nissin who make brakes for Honda bikes, amongst others, do you know Matt?
..

Nice picture Matt.

Clunes, "so it's the lack of light that is the cause of noise...the ISO is not causing the noise."

Well, we'll disagree on this as I still think that it's increasing the gain that cases more noise rather than a lack of light as such. Maybe I can convince you...

You can take a shot in the dark at ISO 100 and it won't be noisy unless you do something like or boost the contrast or lighten the shadows but even a shot in good light at higher ISO will include more noise and you'll see it if you look for it so I think that it has to be a function of the amount of gain rather than simply a lack of photons. By increasing the gain you are introducing more random elements into the signal which would otherwise not be present and speckles are the result whereas if there was no increased gain and few photons there'd simply be a faint or low contrast image or none at all and just blackness.

There are other problems with using higher ISO as you get a reduced dynamic range although this will be off set to a degree or maybe completely, I don't know, by exposing to the right. As they say...

"A 12 bit image is capable of recording 4,096 (2^12) discrete tonal values. One would think that therefore each F/Stop of the 5 stop range would be able to record some 850 (4096 / 5) of these steps. But, alas, this is not the case. The way that it really works is that the first (brightest) stop's worth of data contains 2048 of these steps fully half of those available."

But exposing to the right has to be followed by correction post capture otherwise the result is too bright. The effect seems to be a higher DR image than I'd expect at higher ISO and lower noise. After trying it all I can say is that in practice it seems to work.
Clunes

Don't get me wrong ETTR definitely works and is a good way of maximising the DR as well.

You can (sort of) 'underexpose' and bring back in pp as well if shooting at constant ISO rather than boosting ISO and get the same 'noise' but have more control over how the gain is applied to different areas of an image.

Of course, we are discussing a very complicated business here and it pretty much comes down to interpretation of certain aspects but I still maintain that ISO is not the root cause of noise (but can make it more apparent)

E.g. Fix an aperture and shutter speed - take a photo at 100/200/400/800 - but then boost the EV of the lower ISO shots in pp to match the exposures and the noise will be very similar. This is not what we do in real life of course..in low light we boost ISO to lower shutter speed to stop blur etc

(Oh and taking a shot in the dark at ISO 100 might be very noisy if you let the camera expose 'correctly' as the shutter speed will be very long and a bunch of crappiness will start to appear :) )

Turn the volume up on a crackly record and you hear the 'crackles and fuzziness' more than at a lower volume but it is not the volume itself causing those crackles - merely amplifying them!
Matt

"him" wrote:
^^^ Does the Di28 work with the Canon G6 too?


It works with Canon ETTL 1&2 and iTTL if that helps. Guide number is 28 at ISO200 and compared to the built-in G10 flash, which is limited to three levels of output, this can be finely tuned through the camera's menu.
..

Clunes, I'm still not convinced.

Crackles on a record are essentially no different to the music as they are caused by bumps and pits in the vinyl just like the music is. A better analogy would surely be hum or white noise. Imagine a record or CD with no imperfections to cause crackles or anything else but with only very quiet music. If you turn up the volume too high the white noise generated by the amplifier will be much more apparent and in fact will not have been there in any significant quantity at all before you turned the volume up to 11.

Noise in cameras is expressed as speckles which are not simply amplifications of something that is there but completely artificial artefacts that are not part of the real scene. You can test this by looking at the pattern or possibly by taking a black frame shot. In fact as far as I know black frame shot subtraction is used to remove noise and you couldn't do that if what you were doing was amplifying some hardly noticed but true element of the scene it'd be different for every shot.

The following are both a "black" object photographed with the same aperture and shutter speed settings. Only the ISO setting varies. I think that this tells me that increasing the ISO leads to more noise.





The noise is not just hot pixels, there are coloured ones too, hot ones tend to be white.
..

Mat, I want one.
Clunes

Increasing ISO can lead to more noise as it increases the gain reducing SNR - you said that yourself!

But it is not the CAUSE of the noise.
Clunes

Here's a more real world example of what I mean - sorry they are a bit large :)

ISO1


ISO2


Same Shutter, Same f-stop - different ISO but with exposures brought to match i.e. the 'same' photo at different ISO

I'm in no way saying you can't see differences between the two but purely on a 'noise' basis there is not 'that' much in it...
DaveGibson

Clunes wrote:
.............

ISO1


ISO2

..........

They look identical to me but I have to say that the legend "This photo is currently unavailable" would be more legible if it were bigger.
Clunes

Hmmm - tis showing on mine!!

Will try again 2moro - getting WAY too late although here is an example of what I mean:

http://lagemaat.blogspot.com/2009/08/high-iso-is-bad-right.html
..

Maybe it's just because it's late but that just doesn't make sense to me at all. The guy boosts the low ISO shot 4 stops in post capture to match ISO 1600 and of course he'll get a mess. This is something that you'd never do except in emergencies when quality isn't a concern. I use ISO 3200 with the knowledge that it's a fiddled ISO 1600 but at least that's only one stop, not four. Then he says that a correct ISO 200 exposure will have less noise and that if you care about noise and have enough time you should use the lowest ISO.

The shots I took, soz, can't see yours, were at the same aperture and shutter so the light hitting the sensor was the same and yet the ISO 100 shot has no visible noise at 100% and the high ISO shot is noisy as hell. That makes sense to me.

The only sense I can see in the article is that if you meter for a perfect exposure and then up the ISO and then back it off again post capture you'll effectively do the same as adding exposure compensation to expose to the right but you'd have to back off post capture to get a nice result, not boost 4 stops, that's madness, one stop if you can get away with it will do nicely but two would probably more often than not be a stop too far.

I can understand that a low ISO shot will possibly lead to higher noise as it takes longer to take but a higher ISO shot means more amplification and a drop in DR. You can only degrade with amplification. I suspect that some newer cameras are maybe set up with a bias towards higher ISO performance at the expense of lower ISO performance possibly in the hope that lower ISO shots will mask noise as long as the exposure is good but without such wizardry for marketing purposes signal amplification can only degrade and move things away from reality.

I'll read it again at the weekend and see if it sinks in but at the mo I think it's a theory and an interesting one but not something that I'd want to follow by shooting everything at ISO 1600.


       The Motor Forum Forum Index -> Photography
Page 1 of 1
Create your own free forum | Buy a domain to use with your forum