BeeMan458 wrote:
Forgive my for not quite understanding but if you have 1/3 stop steps and you have twenty-one steps, doesn't that translate out to seven stops?
Your clarification on my above misunderstanding would be helpful.
No picture available of the 31-step 1/3 stop target - the one pictured is a 21-step 1/2 stop target.
That explains the confusion. I thought my monitor had broken or something cause it's calibrated and I was down to only six and two thirds stops (19 of 21 as 20 and 21 run together into pretty much one, unless you angle around and I consider that cheating). Scared me.
Paul -
Your BW cathedral shot is glorious. I miss the days when back in NH in the 70's we had our lab and used Zone XI papers (Brilliant, etc.). It's nice to see the work of an acknowledged master and although we must keep tonque in cheek when reading some of the posts made by forum regulars well known for behavioral problems, it really is refreshing to see work that is informed by the old school and methodically updated to our current state of the art. Glad to see examples of what you do and your clear headed concise views regarding processing. I laughed at hearing (reading) the word halucinogenic, particularly as used in this context. Recovering from arthroscopic surgery just a day ago, this has made for entertaining reading. Of course, John McFadden is a bully... argh...fuhgeddaboudit...
I don't mind the childish comments (digs-n-jabs) so much as I do mind not being able to share a few of my own (respond in kind) as it's against forum rules to make personal attacks but if anybody wishes to make comment, via PM, where I can openly write and the privacy will be respected, John or Art, I'd be happy to encourage them to stop by for a visit.
With all due respect, absolutely my above is spot on. Why can I say that. Because it's based upon personal experience of many years. How much more on base can one get than their personal experience? One only need watch the histogram while in Photoshop to see these changes taking place. If one is critical of the processes and keep's the color pick close at hand, they can readily see and verify the changes taking place and watch detail being wiped out as each step is enacted.
[- snip -]
"Even your standard 8 bit JPEG can represent a huge dynamic range, you will just have coarser gradations between levels."
[- snip -]
08 bit = 00256 shades
10 bit = 01024 shades
12 bit = 04096 shades
14 bit = 16384 shades
16 bit = 65536 shades
Thass a whole lotta difference, in my book, in which to work with.
So as to not to be disrespectful, I'll be happy to read what you have to say as to being the last word in your efforts to show me the error of my ways.
Edited by BeeMan458 on Feb 28, 2007 at 09:23 AM GMT...Show more →
The number of shades you have to work with has absolutely nothing to do with the dynamic range at all. You could have a 16-bit file with a dynamic range of 2 EV if you wanted to, you'd just have really fine gradations between levels. Similarly you could have an 8 or less bit file with a dynamic range of 20 EV or whatever it is you wanted to, you'd just have a lot of posterization. The mapping also doesn't have to be linear and is in fact typically not in the case of developing RAW files. These are extreme examples, but the point is that low bit depth does not equate to low dynamic range, it's a function of your input.
Please reread my original comment to your comment.
"I use BreezeBrowser for the reason you mention above. BB hammers the dynamics of a digital file less then the other conversion programs that I've tried. Everything that's done to a file translates into loss of detail. Work flow hammers dynamic range. How you handle contrast hammers dynamic range. Hue and saturation hammers dynamic range."
The dynamic range, after exposure is set by the sensor, as you pointed out, but in post you can expand on the information you have, in effect, expanding the dynamic range of the image capture information in the shadows.
The problem, each process, if not affected properly will cause clipping and have the affect of remove information. The carefully exposed for highlights will blow due to this introduced clipping, there by having the affect of "reducing" the captured dynamic range as you work to expand the compressed information held in the shadows. The fewer bits that you have to work in, the more efficiently this clipping takes place. The more bits you have to work in, the less chance there is of the information being clipped. Example, 1% of 256 Vs 1% of 65K. I'd rather work in a workspace where I chew up 1% of 65k as opposed to 1% of 256.
When working with harsh lighting files, the file is very unforgiving as each step of processing destroys more and more highlight information. It's a very careful dance necessary to preserve information at this level.
Even the simple act of posting online and the compression necessary will clip and destroy the dynamic range of an image. The least amount of prodding will clip the below information and there's little to no chance of saving the below information in a 256 shade, 8 bit workspace. Once the info's clipped, it's "Good night Irene." and the DR of the image is immediately reduced by that amount; not good.
It seems that folks here just want to argue as opposed to openly sharing with each other.
I'll punch out of the thread as I don't think you or the other folks here understand what it is that I'm sharing as I've tried to explain a couple of different ways and I'm not here to force myself or convince anybody of anything they don't want to understand. What I'm sharing, is something that I've been working on and learning about through a great deal of active effort on my part over several years. It's not something I just "thunk" up five minutes ago and I'm running up the flag pole to see if anybody salutes. If you don't believe or understand what it is that I'm sharing, I'm good.
kisielk wrote:
The number of shades you have to work with has absolutely nothing to do with the dynamic range at all. You could have a 16-bit file with a dynamic range of 2 EV if you wanted to...
And, in fact, you could have a 3-bit file with a dynamic range of 16 EV.
Your values would be 000, 001, 010, 100, 011, 110, 101, and 111, i.e. 2^3 = 8 values (as compared with 65536). But if you show it a high enough contrast image, you could place Ev 2 values on 000, Ev 4 on 001, etc all the way up to Ev 16 on 111.
To be sure you'd have a crappy image that was completely posterized. But the dynamic range would be exactly the same as a 32-bit file shot under the same conditions (the difference being that the 32-bit file would have far more photographically meaningful and resolvable detail by virtue of its intermediate shades)..
Well, it seems that you're talking about clipping of information in post processing, the thread was about the dynamic range capabilities of the camera, not what you do with the file in post. I'm not trying to shoot you down, just point out inaccuracies in some of your statements. I am not sure why you are taking it as a personal attack. As you say, we are here to learn and share, and as someone who develops image processing algorithms I don't think I am unqualified in that regard...
DrPablo wrote:
And, in fact, you could have a 3-bit file with a dynamic range of 16 EV.
Your values would be 000, 001, 010, 100, 011, 110, 101, and 111, i.e. 2^3 = 8 values (as compared with 65536). But if you show it a high enough contrast image, you could place Ev 2 values on 000, Ev 4 on 001, etc all the way up to Ev 16 on 111.
To be sure you'd have a crappy image that was completely posterized. But the dynamic range would be exactly the same as a 32-bit file shot under the same conditions (the difference being that the 32-bit file would have far more photographically meaningful and resolvable detail by virtue of its intermediate shades).....Show more →
Paul,
Yes, exactly the point I was trying to make.
Another thing is, some posters seem to imply that downsampling to a lower bit depth would make you lose dynamic range, which is not necessarily the case. It's true if you simply clip some bits off, but you can remap values instead as I mentioned before. This is where the tone curve comes in when developing from raw...
"...some posters seem to imply that downsampling to a lower bit depth would make you lose dynamic range,..."
That would be me and that's not what I'm writing. As I wrote, you don't understand what I'm writing.
I'm writing about expanding the file and the clipping of information. You may not be able to understand what I'm writing of and it's getting frustrating as I fully understand what you're writing of (curves) and I'm not in conflict with what you're writing as you don't seem to be grasping what I'm sharing and it's just spinning out of control as you say it's not personal and then you add insult with you "imply" comment.
Just an aside, I went through the tone curve learning process, as I stated earlier. I watched it destroy information and came to the conclusion that it was a good idea that was still destructive to information so it didn't serve my purposes. I cast tone curves aside a couple years ago for the destructor of information that it was and moved on to try other, less destructive ideas. Been there, done that, it doesn't work for harsh lighting as much as folks want to believe it does.
I don't see harsh lighting images in the different folders here and there's a reason for it.
I've done blending and many other techniques such as shadow and highlight recovery and didn't like what I saw as to the destruction of information but some here are so convinced as to their information, anybody who thinks differently then they, mustn't have a clue and should be insulted. This is common narrow minded behavior.
You and others here do make folks not want to share because of your responses, Why share? This is a question I need to ask myself as you're not able to understand what it is I have to share because you "develops image processing algorithms" and your mind is blocked up by what you know cause you know too much to learn a different way of thought. It's either that or you and others will lash out with uncalled for needling insults because I won't change my original statement to serve your purposes. Heaven forbid that I should stand by my premise as that's now called arguing.
One last time, I'll try and politely punch out of the thread, you can have the last word and I promise not to respond in this thread to your or anybody else's undeclared personal insults.
Yes you can remove information by tone curves, just as with any other operation, but it's not necessarily the case. If used properly you shouldn't be discarding any information that you don't want to.
Unless you develop your RAWs using a completely linear mapping from the sensor information to pixel values in a 16-bit file, you are *always* applying a tone curve when converting from RAW. You may not even realize it. None of the RAW processors I use perform a linear mapping by default because it looks quite unnatural and not what you would expect. Try it in DPP some time and set the mode to "linear" to see what I mean.
I don't have anything against you or your opinions, and I understand what you are trying to say, I just don't think you're right. If you were losing dynamic range through the application of tone curves or the other tools you stated, then you just weren't using them correctly. If you have a different method of processing that you prefer instead, that's fine by me, but saying that these methods somehow intrinsically discard information is simply wrong.
Pondria wrote:
While ago I proposed a method that one can measure the DR easily and consistently w/o any equipment or chart. And FMers volunteered to post their result. The numbers are pretty consistent with DPR, Imatest and digital imaging resource numbers.
Sorry for the repeat posting. No one except me provided any measurement result in numbers. On what basis do some people discount the actual measurement results ? The arrogance is amazing.
DR measured by different people or sites ( including DPreview and Imatest ) with different methods seem to be consistent with each other, right ? DR is not any mythical quality that only artists or experts can perceive. It is a simple quantity that photo diodes can measure. You have millions of them in your camera.
Speaking for myself I never said that I discount the numbers. They are what they are.
But I do I discount the ability of these numbers to predict a camera's performance under ambient lighting of the reported dynamic range -- because that is not what your methodology measured. And as I said before, without a response curve drawn from a standardized dynamic range target, these numbers don't really tell me all that much about actual camera performance, and over how many stops there is photographically meaningful detail.
Now that the smell of cordite has dissipated, can I ask a small, off-topic question ?
Can any of you who use this "Zone" method of exposing an image, point me in the direction of a book or website that explains it's implementation ?
I'm very intersted in how it stacks up against just using the "expose to the right" method.
I really enjoyed reading this thread and, contrary to what Mr Gardner thinks, I DO value his opinions/methods as much as everyone else's. It's all a means to an end.
Sorry for the repeat posting. No one except me provided any measurement result in numbers. On what basis do some people discount the actual measurement results ? The arrogance is amazing.
DR measured by different people or sites ( including DPreview and Imatest ) with different methods seem to be consistent with each other, right ? DR is not any mythical quality that only artists or experts can perceive. It is a simple quantity that photo diodes can measure. You have millions of them in your camera.
Thank you for the information - it is quite impressive how much detail knowledge is present in this thread.
Pondria: thank you for the table. I am very interested in knowing how you measured this. Is there a link to when this table was first posted or how you arrived at these numbers. What did you look at/ look for and how did you come up with the results.
I can agree with your table as the camera bodies I have used/owned in the past seem to coincide quite well with the numbers in the table.
