brainiac wrote:
People who don't know how to interpret the crops that I have posted on this thread should definitely be buying a 450D instead of spending nearly $3K on a 5D2. Would it be better if I chose not to share my findings?
You should always share your findings, unfortunately there will be always, as well, people to misundertand what you tell them.
brainiac wrote:
Why is everyone having hot flushes about this? If you don't like other people looking closely at the performance of tools that they use in their jobs, become a politician and pass a law against it. Otherwise, don't poison what could have been a perfectly reasonable discussion about small but sometimes visible differences between cameras. That goes for all of you who resort to insulting terms like 'pixel peeper' and so on. If information on this thread doesn't interest you, go elsewhere.
Because you make it sound like Canon lied to you. That you bought the 5DII, sold the 1DSIII, then discovered the noise, and hence the quality, was not as Canon stated and feel slighted. Maybe I misread the thread. I apologize if I did. If it's just a simple discussion of camera qualities, then that's great.
brainiac wrote:
Developed with DPP, no exposure adjustment necessary for the look I wanted, and therefore not underexposed. At 200 ISO a £2000 camera ought to be able to produce tidy shadows in this file, with no exposure adjustment.
It would be interesting - and convincing - to see a screen shot of the original histogram of the image when it was imported into your RAW conversion program - e.g. before any adjustments to the RAW defaults were made.
Would it be possible to post this?
EDIT: Oh, wait, if I understand correctly the histogram WAS posted... more below.
Bottom line. While I originally posted a note of concern about the banding of the noise in the original photograph, if the histogram shown earlier in this thread is the same as or close to the original exposure histogram...
... the problem is with the exposure not the camera.
brainiac wrote:
OK - I can see I'm not getting through here. First, HTP was switched off. This is 200 iso. Secondly, the exposure has not been adjusted in post.
Based on my experience shooting the 5D, understanding post-processing approaches pretty well, and printing, you just described the problem.
You would describe your exposure as correct for the look you wanted. I would agree... if you were shooting transparency film.
However, I would describe your exposure as wrong for this scene when shooting digital. I would have increased the exposure considerably. It would have looked wrong on the LCD display. The unaltered imported RAW file would also have looked wrong.
After simple work with levels, curves and so forth...
... the print would have had the same "look" you are going for but the image quality would have been excellent. In particular the "noise problem" would have been a complete non issue in the print. The whole "issue" presented by this image would have been no issue at all.
Digital has the dynamic range that digital has. If you go beyond that at the high luminosity end of this range you blow the highlights and lose detail in such a way that it may not be recoverable. When you push the lower end of the luminosity scale, by definition you reduce the signal to noise ratio of your image and noise can start to become an issue. Banding can also become an issue because you have only a few levels of data to work with.
On an image like this one where dynamic range is rather narrow, the correct exposure increases the signal to noise ratio (e.g. - reduces noise) by "overexposing" by the standard that you might have used with transparency film - but that is the "correct exposure" in this case. (The correct exposure is not always equal to "the exposure in which the luminosity looks a lot like the original scene." Done much night photography? :-)
You raised the question of what to do in a different sort of image in which the dynamic range exceeds that of the camera system? That is a valid question, and it is one that also has to be addressed with any camera, digital or film. As I suspect you know, there are a variety of ways you can choose to handle that scenario:
1. You could choose to blow out highlights. Typically this is not the best approach but there are situations where it could be - for example if the blown highlights are small specular highlights. This approach sacrifices highlight detail in favor of shadow details (and, on digital, lower noise.)
2. You could choose to expose for the highlights and lose shadow detail - and, on digital, also have to think about shadow noise and banding issues. In general, my preference is to choose this over #1, but every shot is different. By careful and informed post-processing (the digital equivalent of the venerable dodging and burning, plus techniques that are only realistic in the digital darkroom) I can usually manage the shadow noise and get sufficient shadow detail for a fine print.
3. In extreme cases you can do what film photographers have done for years when faced with certain scenes with this "problem" - use graduated neutral density filters. Again, I'm sure you already know this, but grad NDs can provide a good way to handle the situation where one portion of the frame (e.g. sky) is "too bright" and another (e.g. foreground) is too dark. There are limits to what you can do with this, but it clearly can be another viable way to deal with this.
4. Depending upon your subject and shooting methods, another excellent technique is to exposure bracket and combine in post. While there are certainly situations in which this is not a viable option, in many others it is the best option.
There is nothing new here. Underexposure creates certain problems. Overexposure creates certain problems. Super wide dynamic range creates serious challenges. If you bring to bear your understanding of the issues and techniques to resolve them you can get an excellent photograph. The specific techniques vary depending on the shot, your goals, and the medium in which you shoot.
For me, the bottom line is that this thread is a classic tempest in a teapot - a lot of fuss about not much. There is nothing new here - it is important to choose the best exposure for creating the image you have in mind, and here simply choosing a different exposure would have accomplished that without the shadow noise in the original post. Photographers have dealt with these types of limitations in their media for the entire existence of photography - part of the "game" is understanding how the process works and then figuring out the best way to deal with it.
However, I would describe your exposure as wrong for this scene when shooting digital. I would have increased the exposure considerably. It would have looked wrong on the LCD display. The unaltered imported RAW file would also have looked wrong.
After simple work with levels, curves and so forth...
... the print would have had the same "look" you are going for but the image quality would have been excellent. In particular the "noise problem" would have been a complete non issue in the print. The whole "issue" presented by this image would have been no issue at all.
Precisely (emphasis mine) This is essentially the Ansel Adams doctrine of getting maximum data into the image record first, then printing the record the way you visualized it.
This is kind of like a sculptor choosing a granite block at the quarry, then taking it home and chiselling away everything that doesn't look like the work he visualized when he chose the block at the quarry.
For me, the bottom line is that this thread is a classic tempest in a teapot - a lot of fuss about not much. There is nothing new here - it is important to choose the best exposure for creating the image you have in mind, and here simply choosing a different exposure would have accomplished that without the shadow noise in the original post. Photographers have dealt with these types of limitations in their media for the entire existence of photography - part of the "game" is understanding how the process works and then figuring out the best way to deal with it....Show more →
I generally agree with this, and would add a bit.
When faced with the loss of shadow detail in film, the better pictoral option was to go all the way and "print it down" to black rather than leave it a dark gray without detail. We see black shadows all the time with our eyes--when the light level is too low to record any detail, our mind interprets it as "black." We never actually perceive a shadow that has no detail, yet is lighter than black. Zero detail = black, as far as our minds are concerned.
In the same situations that film would show us a gray tone without detail, the digital process will show us nothing but the underlying noise...the equivalent of base fog in film. The better course here, too, is to use Curves to print that detail-less shadow on down to black, which takes care of the noise and also gives us a pictoral depiction more agreeable to the eye/mind.
A problem, though, would be a sensor that depicts a true gray tone with discernible noise. There is a difference between a nicely low level of noise that's only visible if there is zero signal compared to a high level of noise that's visible even above significant signal.
danmitchell wrote:
Based on my experience shooting the 5D, understanding post-processing approaches pretty well, and printing, you just described the problem.
You would describe your exposure as correct for the look you wanted. I would agree... if you were shooting transparency film.
However, I would describe your exposure as wrong for this scene when shooting digital. I would have increased the exposure considerably. It would have looked wrong on the LCD display. The unaltered imported RAW file would also have looked wrong.
After simple work with levels, curves and so forth...
... the print would have had the same "look" you are going for but the image quality would have been excellent. In particular the "noise problem" would have been a complete non issue in the print. The whole "issue" presented by this image would have been no issue at all.
Digital has the dynamic range that digital has. If you go beyond that at the high luminosity end of this range you blow the highlights and lose detail in such a way that it may not be recoverable. When you push the lower end of the luminosity scale, by definition you reduce the signal to noise ratio of your image and noise can start to become an issue. Banding can also become an issue because you have only a few levels of data to work with.
On an image like this one where dynamic range is rather narrow, the correct exposure increases the signal to noise ratio (e.g. - reduces noise) by "overexposing" by the standard that you might have used with transparency film - but that is the "correct exposure" in this case. (The correct exposure is not always equal to "the exposure in which the luminosity looks a lot like the original scene." Done much night photography? :-)
You raised the question of what to do in a different sort of image in which the dynamic range exceeds that of the camera system? That is a valid question, and it is one that also has to be addressed with any camera, digital or film. As I suspect you know, there are a variety of ways you can choose to handle that scenario:
1. You could choose to blow out highlights. Typically this is not the best approach but there are situations where it could be - for example if the blown highlights are small specular highlights. This approach sacrifices highlight detail in favor of shadow details (and, on digital, lower noise.)
2. You could choose to expose for the highlights and lose shadow detail - and, on digital, also have to think about shadow noise and banding issues. In general, my preference is to choose this over #1, but every shot is different. By careful and informed post-processing (the digital equivalent of the venerable dodging and burning, plus techniques that are only realistic in the digital darkroom) I can usually manage the shadow noise and get sufficient shadow detail for a fine print.
3. In extreme cases you can do what film photographers have done for years when faced with certain scenes with this "problem" - use graduated neutral density filters. Again, I'm sure you already know this, but grad NDs can provide a good way to handle the situation where one portion of the frame (e.g. sky) is "too bright" and another (e.g. foreground) is too dark. There are limits to what you can do with this, but it clearly can be another viable way to deal with this.
4. Depending upon your subject and shooting methods, another excellent technique is to exposure bracket and combine in post. While there are certainly situations in which this is not a viable option, in many others it is the best option.
There is nothing new here. Underexposure creates certain problems. Overexposure creates certain problems. Super wide dynamic range creates serious challenges. If you bring to bear your understanding of the issues and techniques to resolve them you can get an excellent photograph. The specific techniques vary depending on the shot, your goals, and the medium in which you shoot.
For me, the bottom line is that this thread is a classic tempest in a teapot - a lot of fuss about not much. There is nothing new here - it is important to choose the best exposure for creating the image you have in mind, and here simply choosing a different exposure would have accomplished that without the shadow noise in the original post. Photographers have dealt with these types of limitations in their media for the entire existence of photography - part of the "game" is understanding how the process works and then figuring out the best way to deal with it.
I am afraid I don't agree that the exposure in the OP image is incorrect or misleading in any way. This is a 200 iso image WITH NO EXPOSURE ADJUSTMENT APPLIED, and yet there is a visible criss-cross/banding pattern in flat shadows. All the rest is hot air.
In particular, the suggestion that I should have overexposed this image to make use of the higher levels neglects the fact that this was a very low contrast picture and a more typical picture would not allow one to expose further to the right without blowing highlights. People are going to see criss-cross texture in large prints in fairly common circumstances, and the intrusion of the criss-cross pattern will leave less room for pushing the shadows, which is a vital method for coping with very contrasty scenes.
brainiac wrote:
I am afraid I don't agree that the exposure in the OP image is incorrect or misleading in any way. This is a 200 iso image WITH NO EXPOSURE ADJUSTMENT APPLIED, and yet there is a visible criss-cross/banding pattern in flat shadows. All the rest is hot air.
In particular, the suggestion that I should have overexposed this image to make use of the higher levels neglects the fact that this was a very low contrast picture and a more typical picture would not allow one to expose further to the right without blowing highlights. People are going to see criss-cross texture in large prints in fairly common circumstances, and the intrusion of the criss-cross pattern will leave less room for pushing the shadows, which is a vital method for coping with very contrasty scenes....Show more →
I know my reply was long - maybe too long - but I addressed both of those issues: why this is not the "right" (or at least optimum) exposure for this scene and how to handle the high dynamic range scenes you mention.
If your exposure choice resulted in "criss-cross" patterns and noise in the print, and a different exposure would have eliminated this without detrimental effects on the final image... you chose the wrong exposure. That seems painfully obvious to me in the context of the sample image in this thread.
brainiac wrote:
In particular, the suggestion that I should have overexposed this image to make use of the higher levels neglects the fact that this was a very low contrast picture and a more typical picture would not allow one to expose further to the right without blowing highlights. People are going to see criss-cross texture in large prints in fairly common circumstances, and the intrusion of the criss-cross pattern will leave less room for pushing the shadows, which is a vital method for coping with very contrasty scenes.
Richard, I've already made a number of large prints from my 5D2, in fact a couple are hanging in a local art gallery at the moment. And there's no "criss cross textures" or any other such imperfections in any of mine, in the shadow areas or otherwise. Perhaps your 5D2 is due for a firmware upgrade, or trip to Canon service for warranty work.
RDKirk wrote:
When faced with the loss of shadow detail in film, the better pictoral option was to go all the way and "print it down" to black rather than leave it a dark gray without detail. We see black shadows all the time with our eyes--when the light level is too low to record any detail, our mind interprets it as "black." We never actually perceive a shadow that has no detail, yet is lighter than black. Zero detail = black, as far as our minds are concerned.
In the same situations that film would show us a gray tone without detail, the digital process will show us nothing but the underlying noise...the equivalent of base fog in film. The better course here, too, is to use Curves to print that detail-less shadow on down to black, which takes care of the noise and also gives us a pictoral depiction more agreeable to the eye/mind.
A problem, though, would be a sensor that depicts a true gray tone with discernible noise. There is a difference between a nicely low level of noise that's only visible if there is zero signal compared to a high level of noise that's visible even above significant signal....Show more →
I'm with you on this, more or less. Your Ansel Adams analogy is right on. I have a hunch he might have a good laugh over this thread...
Here's how I handle extremely wide dynamic range scenes in which I want to retain some suggestion of shadow detail yet minimize banding and noise. You'd approach it differently with the shadow issues you'd have with film, but I'm working with a digital capture. Any or all of the following might be useful on a given image:
1. Apply some noise reduction selectively to the dark area with the noise issue. My goal would not be to eliminate noise (that is an impossibility) but to reduce its visibility in the dark area. Fortunately, there isn't going to be a lot of significant detail in shadows this dark, so one can be pretty aggressive if necessary.
2. Steepen the luminosity curve at the dark end of the curve after grabbing this area with a selection. I want to get black all the way to black yet I want a bit of non-black to remain in many cases - sometimes the "suggestion" of non black tones is plenty. I lock down the middle and lower middle portions of the curve and play with the low end.
3. Using some intentional blur on these areas can also work nicely.
There are also some interesting games you can play with different types of layers: lighten, darken, screen, etc.
3. I'll test this on a print. Quite often stuff that you can see at 100% on the screen is just plain insignificant if not invisible in a print. If it prints well, it is fine with me.
Brainiac - I may have missed it, but have you tried other raw converters to see if that makes a difference? Is it possible the pattern noise and pink blobbing (which I must say, is hardly apparent to my eyes) is more a function of the raw converters than the sensor? Maybe someone didn't spend quite enough time optimizing the raw conversion?
I haven't found any objectional noise at low isos in my 5D2, even when pushed by a few stops. Certainly nothing that would impact a print. I'm using ACR. If anything I find the noise isn't as smooth as I'd like it, and has more of an artifacty feel, but that's only apparent in smooth gradients pushed several stops..
I'd second trying a different raw converter. DXO Optics Pro is especially nice, does a great job with 5DII files, and also have superb noise reduction if/when you want it. What's great about the DXO noise reduction is that it's key to each specific camera and iso, and its applied to the raw data before it's demosaiced. I haven't use it for iso 200 files, but I did use it for iso 6400 files on the 5D II; on those files I usually put the luminosity noise reduction at about 1/2 of the default.
danmitchell wrote:
I know my reply was long - maybe too long - but I addressed both of those issues: why this is not the "right" (or at least optimum) exposure for this scene and how to handle the high dynamic range scenes you mention.
If your exposure choice resulted in "criss-cross" patterns and noise in the print, and a different exposure would have eliminated this without detrimental effects on the final image... you chose the wrong exposure. That seems painfully obvious to me in the context of the sample image in this thread.
I haven't made myself clear. I totally agree with you that THIS SCENE could be exposed more to the right for a better result. That is not relevant to the fact that the general scene, which is what I was testing with this shot, will generally show crusty shadows at 200 iso, as this shot clearly demonstrates. Forget about processing, and just remember that whenever this camera is set to 200 iso and the exposure is not adjusted in processing (i.e. exposure is correct), and there are flat not quite black shadows, a crusty texture will show. I do not expect to have to do half an hour's fiddling with a 200 iso file to make noise invisible in the half shadows.
Please understand that I am not being deliberately obtuse; I have a genuine technical reason for not exposing this scene as far to the right as I could. The technical reason is this: the general scene (NOT THIS ONE) contains a much wider dynamic range, and therefore, generally, overexposure is not feasible since it blows whites. Flat shadows somewhere above black really do look like this at iso 200 on my 5D, and your suggestions about overexposure and pulling in post are completely irrelevant: http://cyberphotographer.com/5d2/snowythames_topleft.jpg
brainiac wrote:
... the exposure is not adjusted in processing (i.e. exposure is correct)
The fact that the exposure is not adjust in post-processing does not correlate to a determination that the exposure is correct. Where do you get the idea that "correct exposure" means that no post-processing adjustment will be done.
Sorry, but that starting point is a rather unusual point of view, to say the least. You are welcome to hold to it if you like, but by doing so you are making things unnecessarily difficult for yourself.
I still say you are railing against two realities that are known and easily dealt with:
1. The best exposure choice for a shot like this is to change the exposure values so that the histogram is moved considerably to the right. There is no negative affect whatsoever as a result and the noise "problem" goes away.
2. Some scenes other than this one will have more dynamic range than your film/digital camera can handle. The image degrades in various ways at the bright/dark ends of the scale and there are various well-known ways to deal with this.
I've seen effects similar to what we see in the underexposed example - and I'm not going to concede that it was correctly exposed - in images I've shot with my 5D, either due to my error in underexposing or due to the fact that the dynamic range was huge. That's just the way it is - and I can get fine prints from such images in virtually all cases.
danmitchell wrote:
1. The best exposure choice for a shot like this is to change the exposure values so that the histogram is moved considerably to the right. There is no negative affect whatsoever as a result and the noise "problem" goes away.
Dan
This is only true if you view view exposure as a technical decision, rather than an artistic decision. I view it as an artistic decision. "Correct" exposure is the exposure that matches my vision for the shot I'm creating, be it dark, blown out, or anywhere inbetwen.
ChrisDM wrote:
This is only true if you view view exposure as a technical decision, rather than an artistic decision. I view it as an artistic decision. "Correct" exposure is the exposure that matches my vision for the shot I'm creating, be it dark, blown out, or anywhere inbetwen.
That would be quite true if we were dealing with transparency material--when the exposed material in the camera is actually the final image media. But not when when the exposure is made on an intermediate recording media. As with negative film, exposure is a technical decision...the exposure that records the greatest amount of the scene's information.
RDKirk wrote:
That would be quite true if we were dealing with transparency material--when the exposed material in the camera is actually the final image media. But not when when the exposure is made on an intermediate recording media. As with negative film, exposure is a technical decision...the exposure that records the greatest amount of the scene's information.
You and Dan seem to understand this concept well.........but why do so many excellent photographers just not get it.....it is really quite simple.........they just can`t seem to brake away from the fact that a sensor is different then a transparency or film . A different medium....a different technique......similar results .................Peter
RDKirk wrote:
That would be quite true if we were dealing with transparency material--when the exposed material in the camera is actually the final image media. But not when when the exposure is made on an intermediate recording media. As with negative film, exposure is a technical decision...the exposure that records the greatest amount of the scene's information.
Once again, if you measure the quality of a photograph by the "amount of the scene's information", yes... But I measure the quality of the photograph by the appeal of the light (or lack thereof, if the case may be...) This reminds me of something I notice around here far, far too often in my opinion. Contrary to (apparently) popular belief, there isn't necessarily a direct correlation to a photograph's technical merit and its beauty. Or to put it another way, if in Richard's original photograph, he envisioned a dark window pane as a subject, then the photograph is exposed properly.
ChrisDM wrote:
Once again, if you measure the quality of a photograph by the "amount of the scene's information", yes... But I measure the quality of the photograph by the appeal of the light (or lack thereof, if the case may be...) This reminds me of something I notice around here far, far too often in my opinion. Contrary to (apparently) popular belief, there isn't necessarily a direct correlation to a photograph's technical merit and its beauty. Or to put it another way, if in Richard's original photograph, he envisioned a dark window pane as a subject, then the photograph is exposed properly.
The idea isn`t to change the look you see or want in the image. It is to simply capture all that is possible as cleanly as possible. Then in post to tone it down to what your vision was or is. A camera is just a tool to create your vision. It as you know works very different then your eye.......it is also very different then a film camera although similar. You have to learn new techniques to fit your vision to the medium...............Peter
it's a bit like the B&W zone system, where you want to maximize the dynamic range that your film can capture, by shooting at its optimal exposure. As long as you remember the "look" that you are trying to capture, you can overexpose the scene to some extent without losing the highlights, and bring it back down in post to remedy any noise issues. Shadow areas, be it at ISO 100, or 3200, will show some level of noise. That's just a fact of life. On the other hand, any very well exposed areas will show no noise even at ISO 6400.
The idea isn`t to change the look you see or want in the image. It is to simply capture all that is possible as cleanly as possible. Then in post to tone it down to what your vision was or is. A camera is just a tool to create your vision. It as you know works very different then your eye.......it is also very different then a film camera although similar. You have to learn new techniques to fit your vision to the medium...............Peter