Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: David Watson on December 10, 2011, 08:03:34 am
-
I would be interested to know if better results can be obtained by using a lower ISO in camera and pushing the exposure in ACR or LR. Which one produces a cleaner image and less noise?
Superficially it seems to me that the computer, with is greater processing power and more complex (?) software, ought to be making a better job than relying on the camera's limited processing capacity.
-
Hi!
On some cameras pushing ISO helps by increasing amplification before feeding the signal to the ADCs. This is pure amplification and it does not involve processing. All Canon models have this behavior. Cameras based on the latest CMOS-designs from Sony will probably benefit little as they don't have pre-amps.
The processing ability in the camera is quite impressive, BTW. My MacPro consuming 200 W with four cores can convert one image/second using all eight thread. My Sony Alpha converts 5 image/second and operates on a small battery.
Best regards
Erik
I would be interested to know if better results can be obtained by using a lower ISO in camera and pushing the exposure in ACR or LR. Which one produces a cleaner image and less noise?
Superficially it seems to me that the computer, with is greater processing power and more complex (?) software, ought to be making a better job than relying on the camera's limited processing capacity.
-
I would be interested to know if better results can be obtained by using a lower ISO in camera and pushing the exposure in ACR or LR. Which one produces a cleaner image and less noise?
From your question I assume that (for whatever reason) you do not have the option of resolving the under-exposure problem using shutter speed and/or aperture. To expand on Erik's reply, increasing the ISO up to about 1600 (in many but not all) cameras will reduce noise and thus allow a cleaner image. Note that doing so comes at the price of reducing the dynamic range (see the graph in Erik's post).
-
I would be interested to know if better results can be obtained by using a lower ISO in camera and pushing the exposure in ACR or LR. Which one produces a cleaner image and less noise?
The one possible advantage I have heard of for the more complicated route of using a lower ISO speed setting and then compensating in post-processing is avoiding blown-out highlights in scenes that have highlights well above your intended "midtone placement".
The two possible advantages of setting the ISO speed at the level appropriate for your intended tonal placement are (1) extra analog amplification before ADC, which can improve shadow noise, and (2) avoiding that extra post-processing step of pushing.
By the way, it seems common that cameras to some extent avoid applying excessive analog gain in producing RAW data anyway: beyond a certain level, higher ISO settings do not increase analog gain, but only (a) apply further digital domain amplification during in-camera RAW conversion, (which is trivial; just bit-shifting), and (b) put information into RAW files suggesting that RAW conversion software on the computer do likewise.
But evidence in recent threads indicates that advantage (1) only applies with some cameras, like Canons (and some older models with inadequate ADC's), and even then only up to about two or three stops above the base ISO speed, IIRC.
So unless a scene gives reason to worry about extreme highlights (liked dimly lit interiors with brightly lit windows, or other mixtures of substantial shadows with brightly lit parts), you might as well do the simple thing: adjust the ISO speed so that the levels come out as you want them without need for "digital push processing".
-
... increasing the ISO up to about 1600 (in many but not all) cameras will reduce noise and thus allow a cleaner image. Note that doing so comes at the price of reducing the dynamic range (see the graph in Erik's post).
Higher ISO speed setting will not decrease dynamic range so long as it does not cause clipping of highlights; so long as the highlights stay in range, reducing the noise floor will indeed increase dynamic range, in the practical sense of range from highest recorded signal level to shadow noise level.
-
Superficially it seems to me that the computer, with is greater processing power and more complex (?) software, ought to be making a better job than relying on the camera's limited processing capacity.
The impressive processing power needed to push ISO in software is as follows:
Let 1067 be a given RAW level in some pixel as captured by the camera. Now the user pushes exposure by 1 stop in his RAW developer. The computer makes the following ultrasophisticated calculus: 1067 * 2 equals 2134. End.
-
The impressive processing power needed to push ISO in software is as follows:
Let 1067 be a given RAW level in some pixel as captured by the camera. Now the user pushes exposure by 1 stop in his RAW developer. The computer makes the following ultrasophisticated calculus: 1067 * 2 equals 2134. End.
except ACR/LR do not work w/ raw data when you operate any sliders in their UI - they convert immediately to prophoto colorspace w/ gamma 1 and push will be w/ that data.
-
Higher ISO speed setting will not decrease dynamic range so long as it does not cause clipping of highlights; so long as the highlights stay in range, reducing the noise floor will indeed increase dynamic range, in the practical sense of range from highest recorded signal level to shadow noise level.
Based on my (admittedly limited) understanding, reducing the noise floor could theoretically increase dynamic range only if the highlight clipping point is not reduced to the same extent. I doubt this is ever the case. As long as the highlights remain within range, increasing the ISO is effective in reducing the noise floor, but it simultaneously lowers (by as much or more) the point at which the highlights will no longer be in range. That is my concept of dynamic range.
-
Mouse,
Let us not debate the definition of dynamic range (which in photography is often mangled relative to its well-established engineering meaning) and just look at image quality results. So long as the ISO speed is low enough that no pixel suffers from clipping, then the upper end of the range is in practice at exactly the same level, but the noise floor is lower, an undeniable improvement with no downsidemfor this particular image. Any unused highlight headroom above the brightest highlights is irrelevant to the quality of this particular image.
The difference is like measuring something that you know is about eight inches long with a yard-stick instead of a foot-ruler and claiming an advantage of triple the dynamic range just because the yard-stick has 360 divisions of 1/10 inch compared to 120 such divisions on the foot-ruler. And if the foot-ruler is instead accurate to 1/20 inch (half the measurement noise), its result will be better, though it still has only 240 divisions, so less "length DR".
-
Mouse,
Let us not debate the definition of dynamic range (which in photography is often mangled relative to its well-established engineering meaning) and just look at image quality results. So long as the ISO speed is low enough that no pixel suffers from clipping, then the upper end of the range is in practice at exactly the same level, but the noise floor is lower, an undeniable improvement with no downsidemfor this particular image. Any unused highlight headroom above the brightest highlights is irrelevant to the quality of this particular image.
Well I certainly do not quarrel with your basic conclusion concerning image quality. It simply restates the reasoning that underlies ETTR; increase the exposure short of clipping desired highlights and thereby lift the shadows that much above the noise floor. And if, for whatever reason, you cannot do this with shutter speed and or aperture, you can achieve the same effect by increasing ISO (up to a point, and only with some cameras).
Nevertheless I will stick to my initial concept of dynamic range; the total number of stops (or whatever unit you choose) between the brightest highlight that can be captured and the deepest shadow that registers above the noise floor (how far above is subject to debate). This concept applies regardless of the luminance range of the particular scene I am photographing, which may be less than, equal to or greater than the dynamic range of the camera.
regards/m
-
I will stick to my initial concept of dynamic range; the total number of stops (or whatever unit you choose) between the brightest highlight that can be captured and the deepest shadow that registers above the noise floor (how far above is subject to debate). This concept applies regardless of the luminance range of the particular scene I am photographing, which may be less than, equal to or greater than the dynamic range of the camera.
Both of you are right, but saying different things. Basically JBL is talking about the DR that can be captured in a particular scene as long as highlights are not clipped, even if ISO is pushed. You are talking more generally, about the maximum DR than can be captured for a given ISO in optimal conditions, i.e. when perfect ETTR can be achieved at every ISO.
If you can achieve ETTR at base ISO, that will be the ideal situation where more DR can be captured (cleanest shadows), so base ISO provides max DR. In that context, lower ISOs always permit to capture a higher DR (that is why DxO DR plots are always decreasing).
But if you cannot achieve ETTR at base ISO with your aperture/shutter, i.e. there is still some highlight headroom, you will be able to increase captured DR (more in some cameras, less in others) by pushing ISO until ETTR is achieved. In that situation, pushing ISO can actually increase captured DR (that is why ISO can be useful to the RAW shooter).
Regards
-
If you in one way or another rely on in-camera JPEG, then increasing in-camera ISO makes a lot of sense.
-h
-
The processing ability in the camera is quite impressive, BTW. My MacPro consuming 200 W with four cores can convert one image/second using all eight thread. My Sony Alpha converts 5 image/second and operates on a small battery.
First, you mac probably consumes power that is unrelated to image processing.
Second, the algorithms are different.
Third, in-camera algorithms are probably highly optimized for the platform (perhaps even implemented in hardware). Software running on your mac is probably more general and with many layers of security, generalization etc between the transistors and the image processing algorithm.
Your raw developer may do operations that usually benefit camera files, but have little or no benefit for your Sony Alpha.
But despite my objections, I agree with you, it is impressive what can be done and still be able to take a lot of images before running out of battery. Wonder what they could do if they made a raw-only camera.
-h
-
except ACR/LR do not work w/ raw data when you operate any sliders in their UI - they convert immediately to prophoto colorspace w/ gamma 1 and push will be w/ that data.
LR, yes. ACR, no. ACR uses the colour space (and associated gamma) of whatever is chosen as the output space in the Workflow Options at the bottom of the screen.
What commercial converters do allow you to work directly on the raw data? All are rendered in some form on screen.
-
LR, yes. ACR, no. ACR uses the colour space (and associated gamma) of whatever is chosen as the output space in the Workflow Options at the bottom of the screen.
Whatever "is chosen as the output space" is exactly that - output space - internally it is prophoto and there is some math there in ACR/LR between the raw data and whatever is actually pushed by the exposure slider.
Eric Chan, for example (one of many) = "...the internal reference scene-referred space in Camera Raw & LR is indeed RIMM (ProPhoto linear)..."
For some other non core operations (but not the exposure correction) there might even further (errors, errors) conversion (prophoto/linear -> probably some modded form of the output space selected).
Eric Chan, for example (one of many) = "...The actual color space used for an operation depends on the routine (e.g., noise reduction, clarity, fill light, HSL adjustments, etc.)....", "...ACR/LR does perform internal color transforms as needed to carry out its various image processing routines (e.g., noise reduction, vibrance, etc.). However, this is all done internally and has no real connection to the user-specified color space of the rendered output file (e.g., sRGB, Adobe RGB, etc.)...."
/ added - one more curious detail found = Eric Chan :: "ACR does use L*a*b* for some internal color difference estimates, e.g., for auto-calculated masks." /
but the first thing ACR/LR do - is to get rid of the raw data by converting that to its internal space - once and only once.
What commercial converters do allow you to work directly on the raw data?
Why do you try to limit yourself to commercial software ?
-
Why commercial software? Because that's what the vast majority of people are going to use. Simple. Take commercial out of the question then. What conversion software doesn't render the image on screen according to some colour space and gamma assumptions?
Then why does the on screen rendering change in ACR when the output space is changed in the Workflow Options? If it were solely based on RIMM, it should act like LR and a change in the output space should be just that and not affect what's seen on screen till after the image is opened into PS, shouldn't it?
-
Both of you are right, but saying different things. Basically JBL is talking about the DR that can be captured in a particular scene as long as highlights are not clipped, even if ISO is pushed. You are talking more generally, about the maximum DR than can be captured for a given ISO in optimal conditions, i.e. when perfect ETTR can be achieved at every ISO.
Agreed completely: my only interest in this thread is how to choose settings (ISO speed, aperture, shutter speed, stops of push or pull in post-processing) to get the best results with a particular scene. I would be happy if most photographers stopped using (and more often abusing) electronic engineering jargon like "dynamic range" and instead used established, practical, relevant photographic concepts and measures like subject brightness range, blown/burnt-out highlights, and visible shadow noise. Or otherwise be clear about how a particular jargon term and measurement relates to photographic objectives, not just spec. sheet pissing contests.
The one case I see where it could be better to use a lower ISO speed that adds unused highlight headroom is when there is lots of uncertainty about highlights, such as when you will have to make a sequence of shots with variable lighting and without time to adjust ISO settings or meter for each shot. The good news is that cameras are starting to offer modes where the ISO speed is one of the parameters automatically adjusted on the basis of metering, and sensors are offering that light metering at millions of points in the scene.
-
Why commercial software? Because that's what the vast majority of people are going to use. Simple.
vast majority does not care about in camera ISO vs push in raw converter
Take commercial out of the question then. What conversion software doesn't render the image on screen according to some colour space and gamma assumptions?
no, the note was not about that - the note was the exposure correction in ACR/LR does not operate on the per channel raw data, but instead on the data which is received from the raw data by applying some math first... some raw converters do operate on the raw data in this case - google and you shall find.
Then why does the on screen rendering change in ACR when the output space is changed in the Workflow Options?
because the end result of ACR work is some data in "the output space", so you shall see that difference...
If it were solely based on RIMM, it should act like LR and a change in the output space should be just that and not affect what's seen on screen till after the image is opened into PS, shouldn't it?
I do not own LR - but in LR you shall see that too.
PS:
If
it is not "If" - it is from Eric Chan.
-
The good news is that cameras are starting to offer modes where the ISO speed is one of the parameters automatically adjusted on the basis of metering, and sensors are offering that light metering at millions of points in the scene.
do they know what you are exposing for ?
-
vast majority does not care about in camera ISO vs push in raw converter
Such as? Why be so obtuse with 'google and you shall find'? If you've got something better, then just say what it is. Do all other raw conversion applications work the way you describe? Some? Which? If all or most then why bother with the 'why restrict to commercial software?' comment? It's pretty clear you have an axe to grind against Adobe/Eric. Whatever the reason for that is your issue.
no, the note was not about that - the note was the exposure correction in ACR/LR does not operate on the per channel raw data, but instead on the data which is received from the raw data by applying some math first
Fine. That wasn't completely clear in your statements.
I do not own LR - but in LR you shall see that too.
Don't believe so because the way LR works is that the output colour space is only applied on Export of the image to the rendered file format. There's no ability (as far as I'm aware) to change the onscreen view to something other than Melissa for colour and sRGB for the histogram (I believe that's the way it works) in LR.
-
do they know what you are exposing for ?
I do not know the details, it is just a trend I have noticed, like Panasonic's iAuto mode. But it seems likely that one criterion is avoiding blown highlights, by highlight priority metering, a feature already offered in some m4/3 cameras for example, and another is minimizing motion blur by using high enough shutter speed. The combination would then come close to my ideal for choice of analog gain level.
Anyway, perhaps a better solution in the end will be like Sony's approach of column parallel ADC with perfect matching of analog gain to the ADC's capabilities and no need for variable gain: ISO speed as just a recommendation for digital level adjustment with simple multiplication, or "single ISO speed sensors".
There are hints that Panasonic is using a similar approach in its GH2 sensor; for example, I think I read in an interview that the GH2 sensor produces digital output, not analog output for off-board ADC as with other Panasonic sensors. So maybe this approach will spread through the industry.
-
Such as? Why be so obtuse with 'google and you shall find'? If you've got something better, then just say what it is. Do all other raw conversion applications work the way you describe? Some? Which?
for example = http://www.luminous-landscape.com/forum/index.php?topic=56863.msg460764#msg460764 -> RPP and RT
It's pretty clear you have an axe to grind against Adobe/Eric.
me ? I am just a grateful reader and that design decision was, I guess, done by T. Knoll way before Eric Chan started to work in Adobe Labs... I guess that was necessary for a ~real time UI... they couldn't redemosaick raw data sufficiently fast in real time to reflect end users' every mouse click-n-drag during their senseless play w/ sliders in ACR.
-
for example = http://www.luminous-landscape.com/forum/index.php?topic=56863.msg460764#msg460764 -> RPP and RT
I'd be hard pressed to consider that a "vast majority". RPP is, I believe, a Mac-only option which puts severe limits on its user takeup. Raw Therapee is an open source platform that has limited development (I know a new version was brought out this past summer(?) after several years of no development - and yes, I know the original developer passed away, sadly) and depending on current/future development may not be as up to date for newer cameras. So both are limited in their user takeup. Back to the earlier question: What commercial software acts in the way you describe?
me ? I am just a grateful reader and that design decision was, I guess, done by T. Knoll way before Eric Chan started to work in Adobe Labs... I guess that was necessary for a ~real time UI... they couldn't redemosaick raw data sufficiently fast in real time to reflect end users' every mouse click-n-drag during their senseless play w/ sliders in ACR.
Oh, no, no issues there at all. ::)
-
To answer some questions raised in this thread:
The internal working space of ACR and LR is Referred Input Medium Metric (RIMM), which has ProPhoto RGB primaries with linear gamma. Temporary excursions are made to other color spaces and image decompositions, as needed, to suit the image processing.
In ACR, there are additional options to set the desired output color space (which is used to drive the histogram and on-screen preview image). These color conversions are done at the end of the image processing pipeline, after the UI-driven controls (e.g., Exposure) are done. In LR, there are also such options (e.g., in Export and Web), though they do not drive the displayed histogram, nor the on-screen preview image. The color conversions are handled in exactly the same way.
Note that when you change a slider in ACR/LR (e.g., Exposure), ACR/LR doesn't simply recalculate the new exposure value. It also has to re-evaluate many other things. This is because your adjusted exposure may influence many other image processing parameters (for example, a brightness-dependent color table, brightness-dependent tone or color masks, etc.).
-
The internal working space of ACR and LR is Referred Input Medium Metric (RIMM), which has ProPhoto RGB primaries with linear gamma. Temporary excursions are made to other color spaces and image decompositions, as needed, to suit the image processing.
Eric - what about using ACR to work w/ TIFF and JPG images ? do we still have conversion from their colorspaces to RIMM first (for example JPG in sRGB -> RIMM -> output color space) ? thank you.
-
RPP is, I believe, a Mac-only option which puts severe limits on its user takeup.
yes, "MAC only" - like Aperture (Apple) or Rawdeveloper (Iridient).
however you can run OS X in VmWare or Virtual Box on any PC w/ a proper virtualizaton supporting CPU and as RPP does not need QE/CI (the same, btw, is true for Rawdeveloper) it runs there just fine and moreover its performance in virtualized environment per $1 cost of your hardware is actually better than running it natively on Mac hardware (Hackintosh of course is the most cost effective solution).
plus again, Mac share is bigger among people who are involved in photography either on a pro basis or as amateurs/hobby.
but yes, ACR/LR, are more "average Joe" friendly, no argument here.
Raw Therapee is an open source platform that has limited development (I know a new version was brought out this past summer(?) after several years of no development
on the contrary it is way more active right now vs when it was only Gabor... we even have here at least two people participating in development here - Emil Martinec and Michael Ezra (may be I misspelled the name).
- and yes, I know the original developer passed away, sadly)
are you talking about Gabor Horvath of RT or Gabor Shorr (aka Panoholic) of Rawnalyze ? I did not know that Gabor Horvath died ???
and depending on current/future development may not be as up to date for newer cameras.
RT is as current w/ new cameras as dcraw is = it is up to date (usual suspects like Sigma excluded - but Adobe's support for Sigma is as good as not supported).
So both are limited in their user takeup. Back to the earlier question: What commercial software acts in the way you describe?
no, I do not beat my wife ;D
-
yes, "MAC only" - like Aperture (Apple) or Rawdeveloper (Iridient).
however you can run OS X in VmWare or Virtual Box on any PC w/ a proper virtualizaton supporting CPU
Not a solution I'd venture many, even professional photograpehrs, would utilise. Unlike here, where the proportion is higher, most professional photogs aren't also computer engineers/programmers.
plus again, Mac share is bigger among people who are involved in photography either on a pro basis or as amateurs/hobby.
That was the case in the past. Don't think the same holds true today. Mac is no longer the 'default' platform for graphics/photography.
but yes, ACR/LR, are more "average Joe" friendly, no argument here.
Nice dig at all the professionals who 'make due' with a 'less than adequate' solution.
are you talking about Gabor Horvath of RT or Gabor Shorr (aka Panoholic) of Rawnalyze ? I did not know that Gabor Horvath died ???
Sorry, I got them confused.
RT is as current w/ new cameras as dcraw is = it is up to date (usual suspects like Sigma excluded - but Adobe's support for Sigma is as good as not supported).
Sure I understand dcraw gets updated. But how quickly compared to others? And, in my experience, dcraw isn't as good as ACR/LR or other software platforms. dcraw is what, I believe, Photomatix uses if you feed it raw images and even HDRSoft admits you're better off converting outside Photomatix and feeding TIFF files.
no, I do not beat my wife ;D
I'll take that as saying that no commercial software works the way you describe.
-
Not a solution I'd venture many, even professional photograpehrs, would utilise. Unlike here, where the proportion is higher, most professional photogs aren't also computer engineers/programmers.
sure, I have no doubt that neither you nor many 'd venture... but we digress from internal spaces of ACR/LR and from what exposure slider is actually doing in various raw converters ;D...
That was the case in the past. Don't think the same holds true today. Mac is no longer the 'default' platform for graphics/photography.
I did not say 'default' platform... my point is that the marketshare of Mac is way bigger in graphics/photography (and if you go up the ladder it becomes more Mac and less Win - and note that I am not using any Mac hardware myself - just OSX/RPP), for example - take P1 U2U forums and compare the activity of C1/Mac vs C1/Win sections - it is around 50/50... but we digress from internal spaces of ACR/LR and from what exposure slider is actually doing in various raw converters ;D...
Sure I understand dcraw gets updated. But how quickly compared to others?
as often as ACR/LR.
the problem is not how soon you can open the raw file (genuinely new formats are very rare thing) - the problem is how good is your camera profile (.icc/.icm or .dcp) available (from OEM or supplied w/ a raw converter or from a 3rd party or self built).
And, in my experience, dcraw isn't as good as ACR/LR or other software platforms.
dcraw (or some refactored libraries like libraw built from dcraw information) is a basis on top of which some raw converters are build (it is used to get information from raw files and as a source of matrix profiles sometimes) but they might (and do) use their own demosaicking, denoising, etc... so a proper raw converter built using dcraw is not just a gui over dcraw.
I'll take that as saying that no commercial software works the way you describe.
no, however RPP (and RT) are used by commercial photographers... it is like denigrate Argyll because it is a free software :-)
-
I sometimes wonder if someone will replace David Coffin if he decides on day (or is forced for some reason) to stop developing DCRAW. He is a very entushiastic person, and he loves what he does late night in his room everytime a new camera format appears. But he is a single person, and I doubt there are many individuals out there capable of (or willing) doing what he does.
-
But he is a single person, and I doubt there are many individuals out there capable of (or willing) doing what he does.
I think you forgot about the authors of libraw.
Joofa
-
I think you forgot about the authors of libraw.
for now libraw still depends on dcraw as a source of information about new cameras (not that there some big changes in a source code if you compare sequential dcraw releases)
-
for now libraw still depends on dcraw as a source of information about new cameras (not that there some big changes in a source code if you compare sequential dcraw releases)
True, however, I was not answering the dependency part but the "capability" needed to contribute to dcraw as asserted by GL.
Joofa
-
Amazing that we customers put up with Camera manufacturers that do not document the file format that will allow us to read our own negatives in the future.
One would think that Canikon could spare a few hours of developer time to make available an example implementation/SDK that decompressed/read the data from each of their raw formats, and presented it as human readable tagged data. As long as no actual image processing was carried out, it should be a lot simpler for them that for someone without access to internal documentation/human resources.
-h
-
Hi,
Statistics on my website indicate both Windows and Mac being about 45% of the visitors. I guess most visitors I have are photo geeks.
The way I see it, users choose the operating system that best suit their needs. For me it was the only reason I switched from Linux to Windows and than to Mac.
Vendors support the OS-es of their choice. That choice may depend own the developers competence. It's not as easy to change OS as to change shirt. There are application frameworks that can be used on multiple OS-es line Qt. Bibble Pro uses Qt and is available on Windows, Linux and Mac. But even using a library like Qt you need to put up with idiosyncrasies of different platforms for color management, sound and so on.
Best regards
Erik
That was the case in the past. Don't think the same holds true today. Mac is no longer the 'default' platform for graphics/photography.
-
Interesting! on your stats iPhone and Android are both equal. Has the expected crossover in the smartphone industry already come?
-
Eric - what about using ACR to work w/ TIFF and JPG images ? do we still have conversion from their colorspaces to RIMM first (for example JPG in sRGB -> RIMM -> output color space) ? thank you.
Yes, the source color space of the output-referred image (TIFF, JPEG, PSD, etc.) will be converted to RIMM temporarily for the purposes of adjustments in ACR/LR. For example, if you started from an in-camera JPEG shot in sRGB, the conversion would be done from sRGB to RIMM.