Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: flyingpanther on August 14, 2007, 10:13:57 am
-
Hello everyone,
I am new to this forum. I recently created a tutorial on how I create my high dynamic range images if anyone is interested. Here is the link: High Dynamic Range Images (http://flyingpanther.wordpress.com/hdr-tutorial/)
(http://farm2.static.flickr.com/1076/1104966677_64b67f7356.jpg) (http://flyingpanther.wordpress.com/hdr-tutorial/)
Tyler
-
Hello everyone,
I am new to this forum.
[a href=\"index.php?act=findpost&pid=133189\"][{POST_SNAPBACK}][/a]
Uh huh...and it sounds presumptuous for one of your FIRST posts to be one that drivers people to YOUR web site. Ya might try to be just a little less blatant about it mate...
It's one thing to mention the tutorial in the course of a discussion about HDR, but it's pure advertising to make a post like this.
-
Uh huh...and it sounds presumptuous for one of your FIRST posts to be one that drivers people to YOUR web site. Ya might try to be just a little less blatant about it mate...
It's one thing to mention the tutorial in the course of a discussion about HDR, but it's pure advertising to make a post like this.
[a href=\"index.php?act=findpost&pid=133246\"][{POST_SNAPBACK}][/a]
Schewe,
I am sorry that this post rubbed you in such a negative way. My intention with this post is to share my experiences with processing HDR images. I have spent some time developing a tutorial which I feel lays out a simple approach to processing this type of image. I did this because I had a difficult time finding a good tutorial when I was learning the process myself.
Although I am new to this forum I have gleaned much help and info from this and others on the internet and I am thankful for those who have helped me along the way.
Is this not the proper place to post such a thread?
-
I think what Jeff is balking at is that one of your first posts on the forum has the appearance of an advertisement, not a tutorial (your sig has a pointer w/ a discount mentioned if they use your name @ the Photomatix site, so you gain financial benefit for folks buying through you, and the tutorial mentions the same thing). This isn't really the place for advertising....
This is a place for sharing techniques, and such. I think it would generate less ire if you either stated right upfront that you have an arrangement with Photomatix, and that you make money on this, etc, etc - or if you'd left the advertising out of your tutorial and your sig (or at least put the advertising at the bottom of the tutorial).
Ok, that said - you have some wonderful shots that you've shown on the forum and your website. The tutorial isn't any different than other Photomatix related material that I've read, though (including the stuff on their website). If you really wanted to add to what's out there, spending time showing examples of how the different sliders work, and such, would be a big improvement on what's already out there. Most everything says exactly what you say - "just play with them until it looks right"
-
Points taken X-Re and thank you for the suggestions.
-
The point I'm making is that of the 8 posts you've made, starting TODAY, the one thread you STARTED was one in which you directed people to YOUR web site which is considered poor form...this is Michael Reichmann's house...and while he doesn't post strict forum rules (yet) ya don't go on somebody else's forums attatched to their web site to drum up business for YOUR web site, ya know?
Wanna write an HDR tutorial? Ping Michael and see if he will publish it here...or at least have the good graces to mention it in the course of a thread about HDR...such as what you did in "Opinions on HDR, What do people think?". That was proper and would have been enough but then you posted: "Monocrome HDR-Rocky Mountain National Park" and followed that up with "Creating High Dynamic Range Images"...
And seem even more hellbent on self promotion with your signature:
--------------------
I created a tutorial on processing High Dynamic Range Images
http://flyingpanther.wordpress.com (http://flyingpanther.wordpress.com)
All in the first day of posting at the LL...
I'm just saying....
-
Jeff,
Do you really have to crap on everyone? So it's the guy's first post. So what if he's proud of it and wants to share -- it's nice work and it's nice he wants to share FOR FREE. We are all highly opinionated here -- I certainly am. But, I try not crapping on others.
Nemo
-
The point I'm making is that of the 8 posts you've made, starting TODAY, the one thread you STARTED was one in which you directed people to YOUR web site which is considered poor form...this is Michael Reichmann's house...and while he doesn't post strict forum rules (yet) ya don't go on somebody else's forums attatched to their web site to drum up business for YOUR web site, ya know?
Wanna write an HDR tutorial? Ping Michael and see if he will publish it here...or at least have the good graces to mention it in the course of a thread about HDR...such as what you did in "Opinions on HDR, What do people think?". That was proper and would have been enough but then you posted: "Monocrome HDR-Rocky Mountain National Park" and followed that up with "Creating High Dynamic Range Images"...
And seem even more hellbent on self promotion with your signature:
--------------------
I created a tutorial on processing High Dynamic Range Images
http://flyingpanther.wordpress.com (http://flyingpanther.wordpress.com)
All in the first day of posting at the LL...
I'm just saying....
[a href=\"index.php?act=findpost&pid=133307\"][{POST_SNAPBACK}][/a]
Schewe,
To be quite honest I found this forum through a search engine and had no idea it was associated with Michael Reichmann's site. The forum header and registration page did not indicate this. I was under the impression this was a forum to share ones work and/or give and receive tips and critique. I am maybe in the wrong. In retrospect I would have taken another avenue in sharing my tutorial and my deepest appologies for any one who I have offended.
I have not posted it to drum up business for my site but simply to demonstrate or guide people through the process or at least show how the program works. That is all. I truly hope there are no hard feelings.
-
Who cares if it's his first post or his four hundredth post. You've got to start posting somewhere and his images are really excellent and he's providing good information and not asking anything for it. I've now been exposed to software I wasn't aware of previously, and that's a great thing. People need to lighten up a bit or maybe even a lot.
-
To be quite honest I found this forum through a search engine and had no idea it was associated with Michael Reichmann's site. The forum header and registration page did not indicate this.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=133319\")
I would have thought that Luminous Landscape Forum > Equipment & Techniques > Digital Image Processing might have been a clue...plus the link back to [a href=\"http://www.luminous-landscape.com/]The Luminous Landscape[/url] just under the Invision Powerboard logo...plus all the Site & Board Matters sub-forums like: About This Site / Discussions about the Luminous Landscape site and this Board and LL Video Journal & Download Video / Discussion on The Luminous Landscape Video Journal & Download Video
But hey, that's just me...
-
Tyler,
nice pics!! and not just one or two. it's obvious that you've put a lot of time and effort into your photos.
and I'm very impressed with the imagekind gallery/store. I wished there were Asian varieties that I could sign up with.
regards,
Gregory
-
Hello everyone,
I am new to this forum. I recently created a tutorial on how I create my high dynamic range images if anyone is interested.
Tyler
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=133189\")
Tyler,
As to the propriety of your post, I will leave that question to Michael, and I think that others should consider doing the same.
That said, many of your images are striking and well worth a look. The usual criticism of HDR photography is that it can be overdone, giving the images a surrealistic appearance. That was my first impression of your seaside image. On further study, I think that the tone mapping is quite successful and gives the scene an open luminous appearance reminiscent of how the old masters handled tone. For example, here is a comparison to a Peter Paul Rubens scene with similar lighting.
(http://farm2.static.flickr.com/1076/1104966677_64b67f7356.jpg)
(http://www.mlahanas.de/Greeks/LX/Rubens/FallOfPhaeton.jpg)
I can anticipate complaints about my comparison of your work with that of an old master and I do not mean to equate the two, but merely note the similarity of the tone mapping. There are better ways to map tones than with a global S-curve as pointed out on this [a href=\"http://range.wordpress.com/2006/07/15/modern-hdr-photography-a-how-to-or-saturday-morning-relaxation/]Blog[/url] (see section "Old Solutions"). Local adaption and other measures are necessary to simulate the luminance of the original scene in the dynamic range available on the print. HDR tone mapping is in its infancy, and further advances can be expected.
Bill
-
flyingpanther,
I agree with nemophoto, your images are stunning. Well done! Thanks for the info.
I wouldn’t worry too much about self promotion around here. It’s clearly part of the culture. Look around. Some feel it’s only for the anointed few. You see, once you’re part of the priesthood life is good. Pay no attention.
p
-
flyingpanther, your HDR images are stunning. I have often wished I knew how to make HDR images---I've tried and haven't had much success with the software (photoshop hdr feature). remapping the pixels is difficult for me to get the gist of. I found your site helpful. Eleanor
-
Posting protocols aside can the Merge to HDR feature in CS3 'produce' a similar type
of merged image as is referenced in flyingpanthers linked tutorial using Photomatrix.
As I have never created an HDR image I also reread an older tutotial posting by Michael
in this sites archive on Mege to HDR using CS2 before testing out both approaches this evening
and found that the resulting images were miles apart with the Photomatrix solution delivering a
much more visually pleasing result.
As stated,I have zero experience with this feature and was wondering,before blowing tons of
time digging deeper,if others have been able to create an HDR image,to their liking,using
CS3 (which i would prefer to use if possible)
Thanks
Mark
-
Posting protocols aside can the Merge to HDR feature in CS3 'produce' a similar type
of merged image as is referenced in flyingpanthers linked tutorial using Photomatrix.
As I have never created an HDR image I also reread an older tutotial posting by Michael
in this sites archive on Mege to HDR using CS2 before testing out both approaches this evening
and found that the resulting images were miles apart with the Photomatrix solution delivering a
much more visually pleasing result.
As stated,I have zero experience with this feature and was wondering,before blowing tons of
time digging deeper,if others have been able to create an HDR image,to their liking,using
CS3 (which i would prefer to use if possible)
Thanks
Mark
[a href=\"index.php?act=findpost&pid=133561\"][{POST_SNAPBACK}][/a]
Ditto the above - I would really like to see a CS3 tutorial on how to do this.
Its not worth me buying specialist software for HDR imaging - I dont do enough of it. I would like to be able to get a good result in CS3 with it when I do want to do it though.
-
Despite the initial complaints about your first post, I think a nice discussion about HDR can take place. I am interested in high dynamic range scenes; I am not so interested in tone mapping.
I have a question to the forum: why people trend to call "HDR" (High Dynamic Range) to what actually is Tone Mapping? everytime I speak about high dynamic range captures, I cannot refer to the term "HDR" as people quickly think of Photomatix tone mapping, which usually looks unnatural to me. And when they look at my high dynamic range pictures they say: "mmm it doesn't look HDR".
High dynamic range simply means that you captured a wide luminance range of detail in your scene, where the deepest shadows are many f-stops far from the highlights. There is no need of tone mapping to be able to talk about high dynamic range at all, they are linked but different concepts. You can apply tone mapping to a low dynamic range image (for instance Photomatix allows to tone map one single RAW file, which cannot be high dynamic range for today's sensor limitations), and you can have a high dynamic range image without applying any tone maping on it.
Would you think of these pictures being "HDR"? they indeed are, both accounting nearly 13 f-stops of real dynamic range. They are not tone mapped however that's why they don't have the local microconstrast look provided by that technique.
(http://www.guillermoluijk.com/tutorial/histogrammar/resultgamma.jpg)
(http://img66.imageshack.us/img66/2452/f1gv0.jpg)
BTW flyingpanther, in your tutorial you claim: "Remember a single camera shot can hold detail in about a 5 stop exposure range at best."
This is not true in modern cameras. A modest DSLR as the Canon 350D can, if properly used (i.e. exposing to the right), register with a good detail up to 8 f-stops.
-
Despite the initial complaints about your first post, I think a nice discussion about HDR can take place. I am interested in high dynamic range scenes; I am not so interested in tone mapping.
[a href=\"index.php?act=findpost&pid=133585\"][{POST_SNAPBACK}][/a]
Guillermo,
You are correct that HDR and tone-mapping are two separate issues. A HDR image is simply a direct linear mapping of the luminance of the scene. Many real world scenes contain a luminance ratio of 5 orders of magnitude, while photographic reflection prints can display only 2 orders of magnitude. Our screens can do a bit better, perhaps 3 orders of magnitude. Special HDR displays are available, but are currently quite expensive.
To make use of our HDR images, it is necessary to reduce the luminance range in the HDR image to something that can be printed, and this is where tone mapping comes in. The old masters of painting used many tricks to create the impression of a high dynamic range (see the link in my previous post), and local contrast enhancement was one of these tricks. Thus, tone mapping is often necessary in HDR imaging if you want to make a print or view the image on a normal screen.
Bill
-
I don't object to a post referring or linking to a site or a product. (But then, I'm not a formally deputized moderator here.) However, it would be helpful if your post states that the tutorial is Photomatix specific. Your tutorial can also benefit by explaining why you choose Photomatix over many other hdr tools.
Hello everyone,
I am new to this forum. I recently created a tutorial on how I create my high dynamic range images if anyone is interested. Here is the link: High Dynamic Range Images (http://flyingpanther.wordpress.com/hdr-tutorial/)
Tyler
[a href=\"index.php?act=findpost&pid=133189\"][{POST_SNAPBACK}][/a]
-
Would you think of these pictures being "HDR"? they indeed are, both accounting nearly 13 f-stops of real dynamic range. They are not tone mapped however that's why they don't have the local microconstrast look provided by that technique.
(http://www.guillermoluijk.com/tutorial/histogrammar/resultgamma.jpg)
(http://img66.imageshack.us/img66/2452/f1gv0.jpg)
How are the "nearly 13 f-stops of real dynamic range" in these images achieved, and from what source?
BTW flyingpanther, in your tutorial you claim: "Remember a single camera shot can hold detail in about a 5 stop exposure range at best."
This is not true in modern cameras. A modest DSLR as the Canon 350D can, if properly used (i.e. exposing to the right), register with a good detail up to 8 f-stops.
[a href=\"index.php?act=findpost&pid=133585\"][{POST_SNAPBACK}][/a]
I was also lead to believe that a sensor's dynamic range is similar to a slide's, i.e. about 5-stops. Can you elaborate or reference the 8-stops?
-
Thus, tone mapping is often necessary in HDR imaging if you want to make a print or view the image on a normal screen.
[a href=\"index.php?act=findpost&pid=133611\"][{POST_SNAPBACK}][/a]
Well, as soon as we are gamma compensating our images we are rearranging the dynamic range codified into their data to fit in our display, print or visual (eyes) devices. Also when we edit these images using curves to locally increase bright in the shadows or make an overall contrast control we are also tone mapping.
I agree in that this is necessary to be able to enjoy all that DR in the final picture.
In what I disagree is that has become a general agreement that "HDR" means completely tone maped images obtained by using specific HDR purpose software focused on local contrast increase techniques which make them have a very particular and unnatural look. Maybe beautiful (this is always subjective) but unnatural like flyingpanter's sunset.
If I hadn't processed my previous image to lift the shadows, this is what you would get. But did my result above look unnatural? can it be called HDR without aynone complaining?
(http://img66.imageshack.us/img66/8906/f0ty9.jpg)
Regards.
-
How are the "nearly 13 f-stops of real dynamic range" in these images achieved, and from what source?
I was also lead to believe that a sensor's dynamic range is similar to a slide's, i.e. about 5-stops. Can you elaborate or reference the 8-stops?
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=133621\")
I have done some intensive tests about dynamic range on my camera lately, and reached the conclusion that DR is limited by camera noise rather by RAW 12-bit levels. As noise becomes more apparent in the lowest f-stops, it is up to your camera sensor's behaviour in terms of noise how much DR you will achieve.
For the Canon 350D as many as 8 f-stops maximum are usable at ISO100. This is a subjective figure however, in fact I am planning to write a program to measure SNR in an objetive way so that different sensor technologies can be compared in terms of a calculated SNR for each f-stop for instance.
However find here [a href=\"http://www.ojodigital.com/foro/showthread.php?t=143407]350D Dynamic Range[/url] some visual tests on 350D DR and noise. By looking at the grain caused by noise in each f-stop area, you can figure out which of them are usable and what is then the effective DR:
(http://img525.imageshack.us/img525/9052/rango31111piccolorho2.jpg)
(the last f-stop with no noticeable noise is the 8th (strong green).
Regarding the 13 f-stops of the images presented, I used 3 images with different exposures: 0EV (correctly exposed), +3 EV and +6 EV. This extreme overexposure allows to expand, if properly blended, DR by as many f-stops as the used overexposure. In the example: 8 (original 350D limit) + 3 + 3 = 14 f-stops. The scene however was not so wide in terms of DR, look at its logarithmic histogram:
(http://www.guillermoluijk.com/tutorial/histogrammar/histolinlog.gif)
The high captured DR in this image exceeds that anticipated by Ansel Adam's Zone System, which only accounted for 9 possible f-stops filled with image information (zones I to IX). That is why some parts of the captured scene appear in A. Adams' zone 0, while they are not pure black and a lot of texture was captured on them through overexposed multiexposure:
(http://www.guillermoluijk.com/tutorial/histogrammar/zonasluma.gif)
The correspondence in those plots is:
Zone X: Burnt pixels
Zone IX: 0 EV
Zone VIII: -1 EV
Zone VII: -2 EV
...
Zone I: -8 EV
Zone 0: -9 EV to -15EV and Black pixels
-
I have done some intensive tests about dynamic range on my camera lately, and reached the conclusion that DR is limited by camera noise rather by RAW 12-bit levels. As noise becomes more apparent in the lowest f-stops, it is up to your camera sensor's behaviour in terms of noise how much DR you will achieve.
For the Canon 350D as many as 8 f-stops maximum are usable at ISO100. This is a subjective figure however, in fact I am planning to write a program to measure SNR in an objetive way so that different sensor technologies can be compared in terms of a calculated SNR for each f-stop for instance.
However find here 350D Dynamic Range (http://www.ojodigital.com/foro/showthread.php?t=143407) some visual tests on 350D DR and noise. By looking at the grain caused by noise in each f-stop area, you can figure out which of them are usable and what is then the effective DR:
(http://img525.imageshack.us/img525/9052/rango31111piccolorho2.jpg)
(the last f-stop with no noticeable noise is the 8th (strong green).
I'm a film shooter and have just toyed with a digital camera. So I can't comment on how you arrive at your "subjective" 8-stops DR, or how noise enters in the process.
Let me ask the question another way. Take a scene that consists of all 9 zones per AA. On a film camera, I won't be able to capture anything close to all 9 zones on a single slide. Are you suggesting that a digital camera like the 350D can come close in doing so?
Regarding the 13 f-stops of the images presented, I used 3 images with different exposures: 0EV (correctly exposed), +3 EV and +6 EV. This extreme overexposure allows to expand, if properly blended, DR by as many f-stops as the used overexposure. In the example: 8 (original 350D limit) + 3 + 3 = 14 f-stops.
That is how I understand HDR merging works. I misinterpreted your post to mean 13-stops from a single shot. Who would need HDR then?
-
I'm a film shooter and have just toyed with a digital camera. So I can't comment on how you arrive at your "subjective" 8-stops DR, or how noise enters in the process.
Let me ask the question another way. Take a scene that consists of all 9 zones per AA. On a film camera, I won't be able to capture anything close to all 9 zones on a single slide. Are you suggesting that a digital camera like the 350D can come close in doing so?
That is how I understand HDR merging works. I misinterpreted your post to mean 13-stops from a single shot. Who would need HDR then?
[a href=\"index.php?act=findpost&pid=133648\"][{POST_SNAPBACK}][/a]
No no, from a single shot the maximum are 8 f-stops in the 350D, at ISO 100, and if you make sure to completely fill the last f-stop even if the scene has such a luminosity that would make that f-stop be empty in a correctly exposed shot (this is called "Exposure to the right").
The good thing of digital is that if we took care not to blow the highlights we can correct exposure down, and not only this is no problem at all, but improves quality of result in terms of tonal richness and specially signal to noise ratio.
The 8 f-stops figure DR is a very agueable appreciation. I just realised that the first 8 f-stops almost have no noise in them (look at the colour regions, up to the middle green they are clearly defined. From that f-stop on grain is very noticeable.
I plan to measure it again, this time with real calculated SNR values for evey f-stop. By doing that, the declared DR of the camera could be that one that provides a SNR over a certain SNR threshold.
Noise in the shadows is the main factor to limit DR in digital. In fact I would say it's the only one. If noise level becomes close or even higher than signal level, you won't be able to distinguish detail. And no detail means no information properly captured.
-
The 8 f-stops figure DR is a very agueable appreciation. I just realised that the first 8 f-stops almost have no noise in them (look at the colour regions, up to the middle green they are clearly defined. From that f-stop on grain is very noticeable.
I plan to measure it again, this time with real calculated SNR values for evey f-stop. By doing that, the declared DR of the camera could be that one that provides a SNR over a certain SNR threshold.
Noise in the shadows is the main factor to limit DR in digital. In fact I would say it's the only one. If noise level becomes close or even higher than signal level, you won't be able to distinguish detail. And no detail means no information properly captured.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=133653\")
I would not expect the noise to suddenly appear in the darkest f/stop, but to gradually increase in the shadows. Below is a noise analysis for the Nikon D200 based on the test results derived by [a href=\"http://www.clarkvision.com/imagedetail/evaluation-nikon-d200/index.html]Roger Clark.[/url]
The tabulated data represent electrons, with the full well of the camera at about 33,000 electrons. For a Zone system type of analysis, Zone 0 corresponds to full well, and the succeeding darker zones have half the electrons of the preceding. Noise has two main components: photon sampling noise (which is equal to the square root of the number of photons collected) and read noise (which does not vary much with signal). Total noise is also tabulated. Read noise dominates in the shadows and limits the dynamic range of the image. Signal to noise is also given. Contrary to popular belief, the highlights actually have more noise but also a higher signal to noise ratio, so that noise is most perceptible in the shadows.
In practice, performance is not as good as predicted from the model, but as Guillermo notes, dynamic range is limited by the noise floor.
Norman Koren (http://www.normankoren.com/digital_tonality.html) has calculated the bit depth required for various dynamic ranges. For example, a bit depth of 12 in a linearly encoded file would give 9 stops of dynamic range, assuming that the darkest stop would have to have 8 levels to avoid posterization. With a gamma encoded file, fewer bits are needed.
(http://bjanes.smugmug.com/photos/184583922-O.gif)
Bill
-
Well, as soon as we are gamma compensating our images we are rearranging the dynamic range codified into their data to fit in our display, print or visual (eyes) devices. Also when we edit these images using curves to locally increase bright in the shadows or make an overall contrast control we are also tone mapping.
I agree in that this is necessary to be able to enjoy all that DR in the final picture.
In what I disagree is that has become a general agreement that "HDR" means completely tone maped images obtained by using specific HDR purpose software focused on local contrast increase techniques which make them have a very particular and unnatural look. Maybe beautiful (this is always subjective) but unnatural like flyingpanter's sunset.
If I hadn't processed my previous image to lift the shadows, this is what you would get. But did my result above look unnatural? can it be called HDR without aynone complaining?
[a href=\"index.php?act=findpost&pid=133622\"][{POST_SNAPBACK}][/a]
This is an interesting problem, Guillermo. It's true your interior room, after lifting the shadows, does not look unnatural and I think the reason is, you have not exploited the full dynamic range of the image. You've left some dark shadows in the corners, shadows which no doubt do not have any noise (or much noise). However, you can probably achieve a similar effect with a lower DR image by just clipping the shadows.
In a sense, the image is unnatural, but in a way we expect (and accept) of a photograph.
Our eyes (pupils) have a remarkable ability to almost instantaneously adjust to differing lighting conditions as we shift our gaze from one part of a scene to another. Your HDR image of the room interior is esthetic. However, if we were actually there when you took the shots, we would notice much more detail in the shadows, specifically the lower left region.
The HDR process effectively allows the camera to more closely mimic the contraction and dilation of the eye's pupil by applying a more correct exposure to each major section of the scene we are photographing.
I tend to think the so-called unnatural effect of tone mapping is due to this presentation in one glance of a scene which the eye could not take in, in once glance in real life. To get the details in the shadows, we have to direct our gaze at the shadows. Our pupils dilate and we can usually see better detail than a single shot from any camera that also has to include the scene out of the window. However, when we direct our gaze at the shadows, we cannot simultaneously take in the scene through the window. An HDR image can, hence it can sometimes appear unnatural. Just a theory. I could be wrong, or at least over-simplifying.
One exciting feature of CS3 (or at least CS3E which I am now using) is the much improved auto-alignment feature in Merge to HDR. One can now merge hand-held shots, provided the longest exposure is not too slow for a sharp result.
It so happens that I often bracket exposures when I shoot high DR scenes, not for merging purposes if I don't have a tripod, but to ensure I have a shot which is very close to being 'exposed to the right'.
Here is one such shot, shooting against the light and therefore perhaps not particularly interesting. This is the 'normal' exposure which actually was a fairly correct ETTR. This is the one I would use to process this scene.
[attachment=3024:attachment]
CS3E now enables me to use the other 2 shots, one stop under and one stop over, to create an HDR image, albeit with just a 2 stop increase in DR but with an effect I cannot get with just one exposure.
[attachment=3025:attachment]
OK, it's unnatural. Too much local contrast, I think. It's easy to do and I didn't even (deliberately) apply any local contrast enhancement.
-
I would not expect the noise to suddenly appear in the darkest f/stop, but to gradually increase in the shadows.
Of course. That's why I think to measure the effective DR of market cameras and be able to make comparisions, some objetive figure as for instance a SNR threshold (SNR0) must be chosen so we can say: OK, I consider this luminosity range where SNR is greater than SNR0 my effective DR as f-stops into that range were acceptably well captured in terms of presence of noise.
Do you know if there is already some official agreement about how to measure DR between cameras?
-
I tend to think the so-called unnatural effect of tone mapping is due to this presentation in one glance of a scene which the eye could not take in, in once glance in real life. To get the details in the shadows, we have to direct our gaze at the shadows. Our pupils dilate and we can usually see better detail than a single shot from any camera that also has to include the scene out of the window. However, when we direct our gaze at the shadows, we cannot simultaneously take in the scene through the window. An HDR image can, hence it can sometimes appear unnatural. Just a theory. I could be wrong, or at least over-simplifying.
I think the problem is simply that most people trend to take tone mapping beyond the limits (in fact HDR programs usually make difficult to avoid doing so). I know people who master Photomatix, and you look at one of their pictures you'll start to wonder if it is, or it is not tone maped HDR, as they are not clearly tone mapped. But there are few of them, and the usual thing is to find very unnatural results.
I think however that starting from a correctly exposed shot where noise has been eliminated from the deep shadows, gives you more control over the result than any tone maping HDR tool if the goal is to achieve a realistic resulting image.
-
Do you know if there is already some official agreement about how to measure DR between cameras?
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=133692\")
Electronics engineers define dynamic range as described in this article by [a href=\"http://www.photomet.com/library/library_encyclopedia/library_enc_dynamic.php]Roper Scientific[/url]. Roger Clark has described how to convert ADU numbers to electrons on his web site, and I think that you could substitute measured noise in ADUs for electrons. But as you can see, the dynamic range is limited by the full well capacity of the chip and the read noise.
Bill
-
I think however that starting from a correctly exposed shot where noise has been eliminated from the deep shadows, gives you more control over the result than any tone maping HDR tool if the goal is to achieve a realistic resulting image.
[a href=\"index.php?act=findpost&pid=133693\"][{POST_SNAPBACK}][/a]
This is another feature of CS3E. In my example above, the 3 shots were all at f8 and 1/50th, 1/100th and 1/200th.
If I could place myself in the same situation (about 2 years ago), with the knowledge that in 2 years time we would have the features of CS3E, then I could have simply taken a rapid burst of 5 shots at 1/200th and got my 2 stops of extra DR, hand-held as well. This is real progress. I congraqtulate the Adobe team.
Both you and Mark Segal have delved into the complexities of curve adjustments, a field that perhaps needs a whole book devoted to it, like Bruce Fraser's sharpening routines (which I haven't read).
The bottom line as I see it is, if you've got the detail in both highlights and shadows, then the final result depends on your Photoshop skills. If the result looks a bit unnatural, then redo it.
Dynamic range is a hornets' nest. The traditional description of 'so many f stops' is woefully inadequate.
This approach is analagous to claiming that a particular lens has a resolution capability of 200 lp/mm. Sure it has, without reference to contrast.
Likewise with DR. If you describe a particular camera as having a DR of 10 f stops without mentioning resolution, then it's equivalent to reporting that a lens has a resolution of 200 lp/mm without mentioning contrast.
Before I'd had a few glasses of wine at the computer, I thought I might test how my 5D fares with a resolution test chart at various exposures, at ISO 100. You know, to get an idea of how absolute resolution falls off as we descend into the shadows.
But heck, do I really need to do this. Has it not been done before? Can we not predict the results?
-
Ray what do you mean when you speak about "resolution" in the shades while talking about DR? you mean ease to distinguish detail because of noise? or you talk about image resolution itself (pixel definition, sharpness,...), despite the possible amount of noise that will be there masking your useful information?
-
Guillermo, I have to agree that most HDR results look ridiculously over-processed. Recently, I have tried the HDR feature in CS3E, and a trial download of Photomatix, and was seriously disappointed by both. You've mentioned the exposure spacing you used to make your very beautiful interior shots, but I would be very interested to hear how you went about compositing the three frames.
I have recently had much better results using variations on a technique published on the web by Mark Galer, using the "Smart Objects" feature of ACR:
http://www.photoshopsupport.com/photoshop-...ange/index.html (http://www.photoshopsupport.com/photoshop-cs3/hdr-high-dynamic-range/index.html)
Still, your images seem exceptionally "natural" in appearance, so you have obviously given this process considerable thought. Thanks in advance.
-
You've mentioned the exposure spacing you used to make your very beautiful interior shots, but I would be very interested to hear how you went about compositing the three frames.
I have recently had much better results using variations on a technique published on the web by Mark Galer, using the "Smart Objects" feature of ACR:
http://www.photoshopsupport.com/photoshop-...ange/index.html (http://www.photoshopsupport.com/photoshop-cs3/hdr-high-dynamic-range/index.html)
I used a program I am developing, the principle is the simplest: select for every pixel that with the highest SNR in the set of images. Have a look at this here (http://www.guillermoluijk.com/article/nonoise/index_en.htm).
BTW Mark Galer's link sounds really interesting, I will read it more carefully later as it speaks about ghosting in multiexposure shots. I am preparing an algorithm improvement to eliminate ghosting and haven't tried it yet, but I think it will work quite well.
Regards
-
Ray what do you mean when you speak about "resolution" in the shades while talking about DR? you mean ease to distinguish detail because of noise? or you talk about image resolution itself (pixel definition, sharpness,...), despite the possible amount of noise that will be there masking your useful information?
[a href=\"index.php?act=findpost&pid=133767\"][{POST_SNAPBACK}][/a]
Guillermo,
I was thinking of a standard which would cut through subjective variability, such as line pairs per millimetre. If a camera can record 50 lp/mm at a full exposure to the right (and perhaps even at a greater exposure, an overexposure), then how many stops less exposure can we use before that resolution of 50 lp/mm is reduced or obscured by noise.
We might then have a reference point. For example, 5 stops of DR at full resolution; 8 stops at half resolution, whatever.
-
Guillermo,
I was thinking of a standard which would cut through subjective variability, such as line pairs per millimetre. If a camera can record 50 lp/mm at a full exposure to the right (and perhaps even at a greater exposure, an overexposure), then how many stops less exposure can we use before that resolution of 50 lp/mm is reduced or obscured by noise.
We might then have a reference point. For example, 5 stops of DR at full resolution; 8 stops at half resolution, whatever.
[a href=\"index.php?act=findpost&pid=133811\"][{POST_SNAPBACK}][/a]
Oh I see the point now, and I think it's a great approach. So you mean measuring how far we can go in the shadows in terms of lp/mm contrast (i.e. resolution) correctly perceived until noises reduces it to unrecognisable levels. I think this is closer to human eye's perception than just a cold SNR figure.
I'll think about that.
-
Well, as soon as we are gamma compensating our images we are rearranging the dynamic range codified into their data to fit in our display, print or visual (eyes) devices. Also when we edit these images using curves to locally increase bright in the shadows or make an overall contrast control we are also tone mapping.
I agree in that this is necessary to be able to enjoy all that DR in the final picture.
In what I disagree is that has become a general agreement that "HDR" means completely tone maped images obtained by using specific HDR purpose software focused on local contrast increase techniques which make them have a very particular and unnatural look. Maybe beautiful (this is always subjective) but unnatural like flyingpanter's sunset.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=133622\")
HDR tone-mapping techniques seem to be distinguished by including a Selection of pixels, which is either used to build a mask and/or to rule blending. So one could postulate that “HDR” starts when a common sigmoidal tone curve is not enough to transform from a “dark, dull & native” RAW state to a pleasing tonality. Even in case that shadow-noise were eliminated by any means, there’s another issue on the other side of the tonal range: i.e. compression of highlight details through such S-curve.
Below please find a couple of links and articles which I think are in agreement with this categorization. Guess there are many ways to skin the cat. Anyway it should be mentioned that gamma encoding is merely irrelevant because it’s invisible in a color-managed environment. It just shifts RGB numbers but not the assigned Lab values. Perhaps only few of us remember Timo’s linear working spaces .
My 2ct & best regards, Peter
--
Based on one single (Raw) file and conversion:
/> Application of a brightening S-curve + inverted Luminosity mask (for highlights protection):
[a href=\"http://luminous-landscape.com/forum/index.php?showtopic=10765&st=75#]http://luminous-landscape.com/forum/index....ic=10765&st=75#[/url]
/> Using the Shadow/Highlight-tool in Photoshop:
http://www.naturescapes.net/062004/gd0604.htm (http://www.naturescapes.net/062004/gd0604.htm)
/> Using an inverted “mask layer” in Overlay blend mode:
http://www.luminous-landscape.com/tutorial...t_masking.shtml (http://www.luminous-landscape.com/tutorials/contrast_masking.shtml)
/> Applying a darkening Curve + Luminosity mask (for recovery of highlight details):
http://www.goodlight.us/writing/luminosity...itymasks-2.html (http://www.goodlight.us/writing/luminositymasks/luminositymasks-2.html)
Blending exposure variants (+/- EV) or multi-Raw-conversions from one single shot
http://www.luminous-landscape.com:80/tutor...-blending.shtml (http://www.luminous-landscape.com:80/tutorials/digital-blending.shtml)
http://www.tofahrn-foto.de:80/index.php?lg=en&pg=tipps.dri (http://www.tofahrn-foto.de:80/index.php?lg=en&pg=tipps.dri)
http://www.adobe.com/digitalimag/pdfs/phscs2ip_hilight.pdf (http://www.adobe.com/digitalimag/pdfs/phscs2ip_hilight.pdf)
http://photoshopnews.com/2005/05/09/multi-...-smart-objects/ (http://photoshopnews.com/2005/05/09/multi-raw-conversions-using-smart-objects/)
I’d also list Photomatrix as well as Photoshop’s HDR feature here.
-----
-
Oh I see the point now, and I think it's a great approach. So you mean measuring how far we can go in the shadows in terms of lp/mm contrast (i.e. resolution) correctly perceived until noises reduces it to unrecognisable levels. I think this is closer to human eye's perception than just a cold SNR figure.
I'll think about that.
[a href=\"index.php?act=findpost&pid=133822\"][{POST_SNAPBACK}][/a]
I've just done a series of shots of my home-made test chart and I'm not so sure the resolution test resolves this issue. Black and white lines seem fairly resistant to image degradation. A lot of degradation of low contrast detail occurs before lp/mm resolution is noticeably affected.
My impression is, if your standards are clean images from the shadows to the highlights, no significant color noise in the shadows and no significant highlight blowing or color clipping, then a camera such as the Canon 5D has a dynamic range of 5 f stops.
If you're prepared to accept a significant degree of image degradation at each end of the spectrum, on the grounds that some useful information is still present, then we could double that DR figure to 10 stops.
Here are my images at ISO 100 using the 5D, including a comparison of 200% crops which span a 10 stop interval (2 secs to 1/500th).
[attachment=3041:attachment] [attachment=3040:attachment] [attachment=3042:attachment]
[attachment=3043:attachment]
-
How are the "nearly 13 f-stops of real dynamic range" in these images achieved, and from what source?
I was also lead to believe that a sensor's dynamic range is similar to a slide's, i.e. about 5-stops. Can you elaborate or reference the 8-stops?
[a href=\"index.php?act=findpost&pid=133621\"][{POST_SNAPBACK}][/a]
A slide has a far greater dynamic range than 5 stops. Slides are, as far as I know, unsurpassed so far in their dynamic range.
-
Despite the initial complaints about your first post, I think a nice discussion about HDR can take place. I am interested in high dynamic range scenes; I am not so interested in tone mapping.
I have a question to the forum: why people trend to call "HDR" (High Dynamic Range) to what actually is Tone Mapping? everytime I speak about high dynamic range captures, I cannot refer to the term "HDR" as people quickly think of Photomatix tone mapping, which usually looks unnatural to me. And when they look at my high dynamic range pictures they say: "mmm it doesn't look HDR".
High dynamic range simply means that you captured a wide luminance range of detail in your scene, where the deepest shadows are many f-stops far from the highlights. There is no need of tone mapping to be able to talk about high dynamic range at all, they are linked but different concepts. You can apply tone mapping to a low dynamic range image (for instance Photomatix allows to tone map one single RAW file, which cannot be high dynamic range for today's sensor limitations), and you can have a high dynamic range image without applying any tone maping on it.
Would you think of these pictures being "HDR"? they indeed are, both accounting nearly 13 f-stops of real dynamic range. They are not tone mapped however that's why they don't have the local microconstrast look provided by that technique.
(http://www.guillermoluijk.com/tutorial/histogrammar/resultgamma.jpg)
(http://img66.imageshack.us/img66/2452/f1gv0.jpg)
BTW flyingpanther, in your tutorial you claim: "Remember a single camera shot can hold detail in about a 5 stop exposure range at best."
This is not true in modern cameras. A modest DSLR as the Canon 350D can, if properly used (i.e. exposing to the right), register with a good detail up to 8 f-stops.
[a href=\"index.php?act=findpost&pid=133585\"][{POST_SNAPBACK}][/a]
These images are fantastic. None of the strange "backlit" character of usual HDR images.
If you could make a tutorial about this, many forum members would be grateful.
-
A slide has a far greater dynamic range than 5 stops. Slides are, as far as I know, unsurpassed so far in their dynamic range.
Slides capture the least dynamic range of any film type, and most good DSLRs can do better. B&W negative film can capture the greatest subject DR.
-
These images are fantastic. None of the strange "backlit" character of usual HDR images.
If you could make a tutorial about this, many forum members would be grateful.
[a href=\"index.php?act=findpost&pid=134076\"][{POST_SNAPBACK}][/a]
There is no secret about those images, they are simple so noise-free in the shadows (thanks to a severe overexposition of +6EV) that they can be processed with a curve to strongly lift the shadows without any posterization or visible noise.
The only "trick" was to leave the window (in the second image) out of the curve mask layer so that it remains unaltered and it does not blow.
The blending process was done through simple pixel selection with a routine of mine, but most blending software (no tone mapping) will provide similar results.
-
There seems to be varied opionions on DR in terms of f-stops in this thread. What are yours: for slides, DSLRs and B/W negatives?
Slides capture the least dynamic range of any film type, and most good DSLRs can do better. B&W negative film can capture the greatest subject DR.
[a href=\"index.php?act=findpost&pid=134154\"][{POST_SNAPBACK}][/a]
-
There seems to be varied opionions on DR in terms of f-stops in this thread. What are yours: for slides, DSLRs and B/W negatives?
[a href=\"index.php?act=findpost&pid=134310\"][{POST_SNAPBACK}][/a]
I've always understood the DR of slides to be 4-6 stops, negative color film 7-9 stops and B&W film 9-11 stops. But it's always problematic in arriving at a consensus on such figures because there does not seem to be a uniform standard defining a cut-off point with regard to image degradation, which inevitably takes place as one approaches the extremes of both the shadows and the highlights.
-
I've always understood the DR of slides to be 4-6 stops, negative color film 7-9 stops and B&W film 9-11 stops. But it's always problematic in arriving at a consensus on such figures because there does not seem to be a uniform standard defining a cut-off point with regard to image degradation, which inevitably takes place as one approaches the extremes of both the shadows and the highlights.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=134330\")
[a href=\"http://clarkvision.com/imagedetail/dynamicrange2/]Roger Clark[/url] has done a nice analysis of digital (Canon 1D MII--one of the best 35 mm style digitals) as compared to slide and negative film and came up with the following: "Kodak Gold 200, in this test, showed 7 stops of information, Fujichrome Velvia 5 stops, and the Canon 1D Mark II, over 10 stops of information!". This is consistent with Ray's observations.
Although slide film has a lower dynamic range than print film, the contrast is much higher than that of negative film and the D-max of slide film is considerably higher than that of negative film. That characteristic means that even low range scanners can capture the dynamic range of negative film, whereas it takes a high end scanner to extract all the information from slides. Because of the low contrast of negative film, it can be difficult to obtain good gradation in the scan.
Bill
-
the Canon 1D Mark II, over 10 stops of information!". This is consistent with Ray's observations.
Bill, don't you think considering a 10 f-stop DR in a 12-bit RAW based camera is a bit too optimistic? if we assume linear behaviour of the sensor into this usable DR, the entire 10th lowest f-stop would be codified in just 4 different tone non-interpolated levels. I think that's too poor even if noise is not present in such deep shadows, isn't it?
0EV: 2048 levels, 2048..4095
-1EV: 1024 levels, 1024..2047
-2EV: 512 levels, 512..1023
-3EV: 256 levels, 256..511
-4EV: 128 levels, 128..255
-5EV: 64 levels, 64..127
-6EV: 32 levels, 32..63
-7EV: 16 levels, 16..31
-8EV: 8 levels, 8..15
-9EV: 4 levels, 4..7
-10EV: 2 levels, 2..3
-11EV: 1 level, 1
The only thing I can think for reaching such a figure (10 f-stops DR) is to consider a non-linear low end of the response curve of the sensor as a usable range that could include more than 4 effective levels to codify degrees of lightness (of course being these levels "stolen" from upper f-stops, -8EV and up). However the dots in the log-log graphic don't seem to show any non-linearity in the input-output response curve:
(http://clarkvision.com/imagedetail/dynamicrange2/dynamic-range-tfcn-b.gif)
And in any case he is using a 16-bit reference, with which I don't agree as those new levels appearing in the 16-bit range and filling it, are purely interpolated. Real captured levels are in a 12-bit range, and only 4 different of them represent the whole 10th f-stop.
-
Bill, don't you think considering a 10 f-stop DR in a 12-bit RAW based camera is a bit too optimistic? if we assume linear behaviour of the sensor into this usable DR, the entire 10th lowest f-stop would be codified in just 4 different tone non-interpolated levels. I think that's too poor even if noise is not present in such deep shadows, isn't it?
0EV: 2048 levels, 2048..4095
-1EV: 1024 levels, 1024..2047
-2EV: 512 levels, 512..1023
-3EV: 256 levels, 256..511
-4EV: 128 levels, 128..255
-5EV: 64 levels, 64..127
-6EV: 32 levels, 32..63
-7EV: 16 levels, 16..31
-8EV: 8 levels, 8..15
-9EV: 4 levels, 4..7
-10EV: 2 levels, 2..3
-11EV: 1 level, 1
The only thing I can think for reaching such a figure (10 f-stops DR) is to consider a non-linear low end of the response curve of the sensor as a usable range that could include more than 4 effective levels to codify degrees of lightness (of course being these levels "stolen" from upper f-stops, -8EV and up).
And in any case he is using a 16-bit reference, with which I don't agree as those new levels appearing in the 16-bit range and filling it, are purely interpolated. Real captured levels are in a 12-bit range, and only 4 different of them represent the whole 10th f-stop.
[a href=\"index.php?act=findpost&pid=134381\"][{POST_SNAPBACK}][/a]
Guillermo,
As Ray pointed out, calculation of dynamic range of a photograph is not exact, and the floor may be affected by posterization as well as noise. How many levels are needed in the darkest zone is somewhat arbitrary, but 4 levels is probably not sufficient. I think Roger was using a noise floor.
Norman Koren's Imatest allows calculation of dynamic range associated with varying levels of quality. Here is one such test for my Nikon D200 (which has more noise than the Canon 1D M2), and a low quality image with a dynamic range of 10.2 f/stops is obtained. Again, this is using noise as the floor. Norman agrees that the darkest zone should have 8 levels, which would limit DR to 9 stops with a 12 bit image.
(http://bjanes.smugmug.com/photos/66573454-O.png)
I would think that the new Canon 1D MIII with its lower noise and a 14 bit AD converter should be able to get 10 stops of DR easily.
Bill
-
You guys are making me dizzy! Not that its all that unpleasant, sometimes I drink wine or beer to the same effect .
Seriously, very interesting discussion. I really like the results. I am blown away by some of those images. I now realize that it is part of many images I have seen here and elsewhere. Is there a simple tutorial as to how to use the HDR in CS3?? I went to the tutorial link by Mark Galer. Its very impressive but not all that easy to execute as part of work flow. Too many alt-option-del key presses. If I could remember all that I'd have been an astronaut!
-
I just played with the Photomatix software. Pretty amazing program and far outstrips Photoshop HDR. The only way I could initially play with the program was converting a RAW image three times. Not the best but workable. The program very quickly assembled the images and tone mapping did the rest. I had a very credible image without a lot of work. It will be interesting to see what it does when I do a series of exposures intentionally.
Photoshops's HDR conversion took about three times longer and the result was totally unusable. I'm grateful Tyler brought this slick little program to my attention.