Fujifilm X100 -- High ISO pictures

My point is that I think people are expecting this camera to be magic and it will not be. People are complaining about ISO 12800 quality ... a bit ridiculous really. You wouldn't buy the camera because it doesn't perform well over ISO 3200? That seems silly to me. However, it is your money so who am I to diss.

Agreed - those 12800 shots look amazing banding or not. Banding in shadows is cured by making shadows darker. Most cameras don't even go to 12800.
 
Personally, I've never bought a digital camera that I didn't see RAW files from on my own computer screen.

As other posters indicated above, the posted images are far too small for me to draw any conclusions about overall IQ.

I think it's fair to offer a balanced critique of what information/images are available for the X100 so far.

As for IQ at iso12800 - if Fuji didn't intend for it to be used/judged at this ISO then it could have just as easily maxed it out at ISO6400.
 
Last edited:
As for IQ at iso12800 - if Fuji didn't intend for it to be used/judged at this ISO then it could have just as easily maxed it out at ISO6400.

I don't think we should hold the X100's max ISO value against Fuji, even if images aren't ging to be pretty at that value. The max. ISO vaule has for many years been an issue of specmanship in the digital camera business.

Manufacturers make cameras with this sensitivity because it makes for impressive specifications. In all fairness, most manufacturers even concede somewhere in their documentation that this sensitivity value will actually only be a last resort to capture an image irrespective of IQ instead of not being able to do so at all.

It makes much more sense to evaluate IQ at technically meaningful ISO values, which mostly are 1/4 to 1/2 of the maximum. If the X100 is good at such an ISO, it will be a lot more than just good enough.
 
I don't think we should hold the X100's max ISO value against Fuji, even if images aren't ging to be pretty at that value. The max. ISO vaule has for many years been an issue of specmanship in the digital camera business.

Manufacturers make cameras with this sensitivity because it makes for impressive specifications. In all fairness, most manufacturers even concede somewhere in their documentation that this sensitivity value will actually only be a last resort to capture an image irrespective of IQ instead of not being able to do so at all.

It makes much more sense to evaluate IQ at technically meaningful ISO values, which mostly are 1/4 to 1/2 of the maximum. If the X100 is good at such an ISO, it will be a lot more than just good enough.

I completely agree that the X100 shouldn't be measured on high iso performance/IQ alone... in fact, I personally don't care (I'm satisfied with my GF-1 iso at 400 and rarely use it above that)... however, Fuji has to be savvy enough to know that the online pixel peepers will compare it's high ISO performance to other DX or even FX sensors that offer equivalent ISOs...

the X100 is being marketed as being 'different' from the other similarly priced digital cameras and they have certainly taken a novel approach to marketing it... so why fall in line with the high ISO wars like everyone else?
 
People who spend a lot of time working in low light situations (EV 6 or less) will produce much better images using a camera with a 135 format sensor compared to a APS-C camera. In this case the additional cost and extra weight and size are the price one pays for working in these conditions. The X100 will not be the best choice for this type of work.

I must say that my D700 can produce great color 8X12 prints at ISO 1600. One thing is clear though, even with the D700, as ISO goes up the IQ suffers greatly. It's not just the noise, but the dynamic range and contrast suffer as well. Exposure accuracy becomes so much more important above ISO 400.

For my photography I just need color images that print well at ISO 800. My D200s and LUMIX G1 can't/couldn't do this. My D300 and D700 can.

If the X100 RAW files hold up at ISO 800 I will be thrilled. But I would be thrilled if ISO 800 RAWs hold up for any camera I use. For my needs B&W gives me one more stop of EV with the above mentioned cameras. So I would want X100 ISO 1600 B&W images to be printable at 8X12 or maybe even one size larger.

Given current APS-C sensor technology, the X100 should meet my needs. I do not expect it to perform as well as a cameras with 135 format sensors.
 
That output at 12800 is pretty darn good. Not sure if it is actual banding or wood grain or what but still acceptable at that level. If I can get something usable at iso 3200 I will be more than satisfied. The bunny seals the deal.
 
Downscaling averages out the noise, gaining you 1 net stop of ISO for every factor of 2-4 or so (depending on the sensor pattern), but at a heavy cost to resolution.

Interesting figures presentet in an absolute context. Are these figures theroretical/mathematical founded, or are they based on perception - as in blinded "real life" trials? Just curious: sources please?

As for "needing" ISO 6400 or 12800 (or higher), the combination of smal camera/large sensor + high ISO performance = better street camera. If IQ is what we can anticipate so far, the X100 will outperform the M9 in this area (all things being equal - which of course they arnt :=)

Kind Regards
 
Interesting figures presentet in an absolute context. Are these figures theroretical/mathematical founded, or are they based on perception - as in blinded "real life" trials? Just curious: sources please?
You sound slightly suspicious but it seems logical to me. Double the area of a sensor and you will achieve one stop more sensitively, and halve the noise. Halve the size of the eventual photo, as seen, will do the same thing.
 
The more important question is if you're in a situation that you have to use ISO 6400 at F2 or 12800, how would the contrast detection AF system of X100 would perform? Or for that matter how would the MF through the EVF would work in really low light situations?

High ISO performance if not combined with top notch AF system or a large and bright VF that allows one to focus accurately becomes pretty much a means for shooting with faster shutter speeds in normal or low light situations.

I'd be more interested for a top notch base ISO performance for such cameras that have contrast detection AF.
 
Interesting figures presentet in an absolute context. Are these figures theroretical/mathematical founded, or are they based on perception - as in blinded "real life" trials?

Both, sort of. The basis is mathematical, the idea being that noise is more or less random and that because of this randomness, the noise averages out when you put more than one pixel together. We're in a binary environment here, where averaging over 2 neighbouring pixels in each direction statistically cancels out 1 bit of noise, netting you 1 stop of dynamic range. The precise gain depends on the nature of the noise (and probably also the arrangement of coloured pixels on the sensor). In order for this to make a significant impact, you have to add a lot of pixels together, so you lose lots of resolution.

I'm saying "both" (i.e. not only mathematical, but also practical) because it's very common to make use of this to maximize dynamic range in image processing. It's what scanner software does when you do multipass scanning (scan the same pixel multiple times, and for every increase of two in the number of passes you get one extra bit's worth of dynamic range). It's also what people use, for example, when preserving grayscales in downscaling black-and-white images - take an 8-bit picture, convert to 16 bit, downscale (netting you more than 8 bit worth of grays because of the noise being canceled out), and then you have some reserves for adjusting levels, curves etc. before converting the final result back to 8-bit - good for avoiding jagged histograms. So there are a lot of practical applications of this effect.

So looking at heavily downscaled pictures for high ISO performance is misleading, because the scaling hides many of the high ISO artifacts. Converting the whole thing to 8-bit JPEGs (meaning you can get rid of some of those pesky least significant bits where the noise happens) and flattening them a bit with the JPEG compression algorithm also helps.
 
You sound slightly suspicious but it seems logical to me.
Greetings paul. Im allways skeptical when numbers are claimed without reservation. Yes its logical that (apparent) noise is affectet by manipulating the size of the photo.

Double the area of a sensor and you will achieve one stop more sensitively, and halve the noise. Halve the size of the eventual photo, as seen, will do the same thing.
So you are compairing the Sensor physics with the Photo resizing math, claiming there to be a direct linear relasionship - if I understand you rightly? Again Im skeptical, sorry Paul. Sounds much like oranges and apples. Again, sources please?

And is there a linear connection of noise reduction in ones perception (not math/physics), when the photo is halved once again? Remember perception of noise is what we ultimatly talk about here, when judging the afore mentioned coffee picture.

My perception tell me the 12800 ISO is allmost doable, the 6400 is doable and allmost good. Yours maybe different, but I doubt its mathematically linear :=)

Kind Regards

Niels
 
Greetings paul. Im allways skeptical when numbers are claimed without reservation. Yes its logical that (apparent) noise is affectet by manipulating the size of the photo.


So you are compairing the Sensor physics with the Photo resizing math, claiming there to be a direct linear relasionship - if I understand you rightly? Again Im skeptical, sorry Paul. Sounds much like oranges and apples. Again, sources please?

Kind Regards

Niels
I think rxmd has answered this pretty well. In practical terms, too, it's undeniable that if you reduce the size of a photo the noise is less obvious.
 
High ISO performance if not combined with top notch AF system or a large and bright VF that allows one to focus accurately becomes pretty much a means for shooting with faster shutter speeds in normal or low light situations.

Nail, hit on head. Now if Fuji can nail both things, I'll turn my pockets inside out.
 
Greetings rxmd - and thx for your thorough answer :=)

As for the pixel/math question Im not at my best here, but I am familar with the partiel volume effect in digital imaging, when 1 pixel "has do decide" what to show (ie the binary environment). But as you say "The precise gain depends on the nature of the noise" - if the fex noise is defined as grain larger than 1 pixel in the photo, the noise reduction will not fall linearly with photo size reduction - but rather fall rather softly at first and then accelerate with later photo resizing. The actual mathematical algoritm one uses also plays a part. Am I right here?

On perception of images you claim that "I'm saying "both" (i.e. not only mathematical, but also practical)" - and then explains scanning and downsizing tecniques, as examples of perception of images. I have assumed all along that we all understand perception as "human perception" - how you judge a picture, with noise, structures, colors ect. So you really havnt argued for the Real Life scenario here.

My point is that when you judge the coffe photo at the top of this thread, looking for amunt of noise, you dont start all over as this was your first photo (experience is hard to put on math), and you dont only watch for noise in the shadowns - but also for other clues, like the fine micro structures on the red rim of the cup, or of course the structures of the rabbit foam.

I claim that if you do a blinded trial with experienced people you would - to a certain extent - be able to "see through" the compressed image in order to estimate the original noise of the picture. (of course you can fool these people if you wantet to :=)

My perception tell me the 12800 ISO is allmost doable, the 6400 is doable and allmost good. Yours maybe different, but I doubt its mathematically linear :=)


Kind Regards
 
High ISO performance if not combined with top notch AF system or a large and bright VF that allows one to focus accurately becomes pretty much a means for shooting with faster shutter speeds in normal or low light situations.

AF sure...I agree. But I shoot in the dark at 3200 at 1/10 and see perfectly through my cameras LCD. Oh yeah I forgot most people hate LCDs.
 
But as you say "The precise gain depends on the nature of the noise" - if the fex noise is defined as grain larger than 1 pixel in the photo, the noise reduction will not fall linearly with photo size reduction - but rather fall rather softly at first and then accelerate with later photo resizing. The actual mathematical algoritm one uses also plays a part. Am I right here?

Partly. When speaking of the nature of the noise, I wasn't really referring to the size of grains or something - in a digital camera noise quite naturally is spatially limited to the pixel size. I was referring to things like the frequency distribution of the noise.

However, there may be some spatial aspects to digital camera noise, too, such as when you have a very bright pixel, noise in the surrounding pixels might be affected. Those things are difficult to model, but when we are talking about downscaling an image 20 times we can probably mostly ignore them, because I'm not sure how a bright pixel would affect a photosite 20 pixels away (and if it did, there would be major things to worry about regarding image quality of the sensor!)

If you assume something like film grain, which (when talking about digital images) consists of lumps larger than the resolution of the image, the mathematics breaks down and works differently.

Niels said:
On perception of images you claim that "I'm saying "both" (i.e. not only mathematical, but also practical)" - and then explains scanning and downsizing tecniques, as examples of perception of images.

Actually I wasn't referring to the perception of images, but to the mathematics of the scanning and downsizing process. It's not like we can retreat into subjectivity, into "judging" things and into and "perception is different" just because it's pictures we're talking about. The fact that in scanning you gain 1 bit (= 1 stop) worth of useful dynamic range above the noise threshold for every increase in the number of scanning passes by a factor of two, for example, is a very simple mathematical result.

This kind of mathematical result applies even if there is no human looking at the picture at all. You can measure the dynamic range above the noise threshold, or you can take the histogram, do a Fourier transformation of it and see if it's jagged or not, without ever looking at the picture. With downsizing it's basically the same, admittedly with some pitfalls such as those mentioned above, but even those largely aren't perception-related.

I claim that if you do a blinded trial with experienced people you would - to a certain extent - be able to "see through" the compressed image in order to estimate the original noise of the picture. (of course you can fool these people if you wantet to :=)

No, it's not that easy to take people and give them two pictures downscaled 20 times and ask them "which one of them was the noisy one". Once you mash 20 pixels into one, they are gone. There is basically no way of reconstructing them, the "fine microstructures" are gone forever, and the large structures become the new fine microstructures. Otherwise it would be easy to do CSI-style image manipulations where you have a ten-pixel blob through a magical process you arrive at a clear image of a human face.
 
Greetings rxmd :=) I clearly have some catching up to do on the math - fex I never really have understood Fouriers transformation heh.

As I have implied I would like to go the other way round: look at the actural picture, ie the coffee picture, and make an assumtion: is this picture usable to evaluate (among other things) noise, and if yes, how do I grade it.

I will add that I evaluate at least 200.000+ digital pictures pr year and have done so for 12+ years. Not to put myself on a high horse - any schmuck with a pretty CV can be wrong - but to point out that I have some substance in my evaluation. And no Im not 100% sure (cause I havnt got the numbers :=)

May I kindly suggest that you put the numbers aside for a moment or two, use the experience you undoubtly have, and evaluate said photo. I would be most interestet in your thoughts.

My perception tell me the 12800 ISO is allmost doable, the 6400 is doable and allmost good. Yours maybe different, but I doubt its mathematically linear :=)

Kind Regards
 
Back
Top Bottom