Pages: [1]   Go Down

Author Topic: HDR workflow question  (Read 5493 times)

gerafotografija

  • Full Member
  • ***
  • Offline Offline
  • Posts: 224
    • gerafotografija.wordpress.com
HDR workflow question
« on: November 28, 2013, 03:58:41 pm »

I've been starting to notice the image problems in simple automated HDR conversions, and have been trying to put together easy solutions.

In particular, I needed to find a way to correct for the most obvious HDR photomerge artifacts that show up frequently when using hand-held captures with a compact camera.

The main deficiencies I see, at least with something like the Fuji X20 or OMD EM5, are due to slight shifts in alignment from frame-to-frame even at 7-9fps, limited bit-depth (12 bit or less) leading to posterization or clipping, and relatively high base ISO noise levels.

I think I solved at least some of these issues -- here's a link to a short description and example.

I'm curious what approach others take, can anyone comment?
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: HDR workflow question
« Reply #1 on: November 28, 2013, 04:08:16 pm »

Hi,

I have tried HDR a few times, but actually found that is very seldom needed. Today's DSLRs have an incredible dynamic range and HDR is simply not needed. Here are short suggestions, if you use Lightroom or ACR.

1. Use a graduated filter
2. Pull down brightness,  -100 is quite OK
3. Adjust blacks so you get some pure blacks
4. Increase shadows so you get decent shadow detail

Read these two articles:
http://echophoto.dnsalias.net/ekr/index.php/photoarticles/63-lot-of-info-in-a-digital-image
http://echophoto.dnsalias.net/ekr/index.php/photoarticles/61-hdr-tone-mapping-on-ordinary-image

Best regards
Erik

I've been starting to notice the image problems in simple automated HDR conversions, and have been trying to put together easy solutions.

In particular, I needed to find a way to correct for the most obvious HDR photomerge artifacts that show up frequently when using hand-held captures with a compact camera.

The main deficiencies I see, at least with something like the Fuji X20 or OMD EM5, are due to slight shifts in alignment from frame-to-frame even at 7-9fps, limited bit-depth (12 bit or less) leading to posterization or clipping, and relatively high base ISO noise levels.

I think I solved at least some of these issues -- here's a link to a short description and example.

I'm curious what approach others take, can anyone comment?
Logged
Erik Kaffehr
 

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: HDR workflow question
« Reply #2 on: November 28, 2013, 07:43:20 pm »


I think I solved at least some of these issues -- here's a link to a short description and example.

I'm curious what approach others take, can anyone comment?

That linked HDR demo shows the issue we all face concerning which portion of a high contrast image is to be adapted to in emphasizing what's important (from what I see it's the boat in shade). The alternative is to choose to mimic the overall character of light which from that scene depicts extreme differences between the clarity of bright sunlit boats in the background against the darker, low contrast/low clarity of the foreground boat in the shade.

In the last sample in that linked article both extremes in character of light were compromised into one busy looking scene with emphasis to neither.

My suggestion would be for the foreground shade to go a bit darker overall while preserving some clarity/definition (not too much for shade lit by diffused light) and brighten the background while preserving much more clarity by comparison. That's hard to do, see and keep track of in a long editing session stacking bracketed exposures.
Logged

gerafotografija

  • Full Member
  • ***
  • Offline Offline
  • Posts: 224
    • gerafotografija.wordpress.com
Re: HDR workflow question
« Reply #3 on: November 28, 2013, 10:27:41 pm »

Thanks for the thoughtful comments on this topic. I guess the scene is a little upside down from the compositional perspective, with good light in the background rather than foreground, but it makes for an interesting exercise.

I like the idea of emphasizing the difference between foreground and background light while boosting definition and clarity. I'll see if I can push it a little more in that direction.

Regarding the single file approach, I thought the noise level in the shadows and loss of detail in clipped highlights would limit what's possible with a relatively small 12mp 12-bit raw file, but it could be interesting to compare to the image stacking version to see which works better.
Logged

Tony Jay

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2965
Re: HDR workflow question
« Reply #4 on: November 29, 2013, 02:11:32 am »

Everyone has made some useful comments regarding HDR.
I have done several HDR images, usually shot at sunset, or more likely after sunset where the dynamic range really is way beyond what any camera could capture so I do have a little insight.
I would never try to capture images for an HDR image without using a tripod and remote release.
(I even use a programmable remote that automatically changes each capture's exposure for as many captures as is necessary.)
One tends to have enough issues with movement within the scene without complicating the issue further with slight changes in perspective.

Also now that Lr (and ACR) allows manipulation of 32-bit TIFF files I like to do my tonal manipulations in Lr using the usual sliders and tone curve. I aim for an image that would look essentially natural to any experienced photographer apart from more shadow detail and the lack of noise in the shadows.

That said, Eric is right on the money when he comments that HDR is becoming increasingly redundant due to the improvements in sensor quality as regards both dynamic range and noise generation.

Tony Jay
Logged

PhotoEcosse

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 712
Re: HDR workflow question
« Reply #5 on: November 29, 2013, 11:23:01 am »



That said, Eric is right on the money when he comments that HDR is becoming increasingly redundant due to the improvements in sensor quality as regards both dynamic range and noise generation.

Tony Jay

Much of the time, that is correct. However, there are some scenes - particularly those with both snow-capped mountains and water at sunset - where even the 14.3EV sensor of the D800 will produce both blown highlights and solid shadows when the Highlights and Shadows sliders of LR (or ACR) are at their left and right extremities respectively. On occasion I have found it necessary to take 9 exposures at 1-stop intervals to provide sufficient data for HDR processing in Nik HDR EfexPro. (9 exposures because my D800 has only 1-stop increments in auto-bracketing; in practice, I would only use five of the files at 2-stop intervals). But, as has been said, to avoid problems when shooting such a sequence, a really firm tripod is a pre-requisite.
Logged
************************************
"Reality is an illusion caused by lack of alcohol."
Alternatively, "Life begins at the far end of your comfort zone."

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: HDR workflow question
« Reply #6 on: November 29, 2013, 03:08:05 pm »

It's clear some folks here would go through the trouble of using a tripod and stacking bracketed exposures even with a 14+EV DSLR instead of employing both Parametric and Point curve in ACR/LR which is far easier to do with a single exposure.

I've done quite a few single exposure sunset landscapes using a 2006 8EV DSLR recovering enough detail on both ends of the luminance spectrum even on jpegs just exposing for highlights so I'm not getting what folks are wanting to achieve with even better equipment.

Our eyes have to adapt to either the brightly lit or the shade as in the OP's linked boat dock image. Forcing an adaptive appearance in post for both dark shadows and highlights will certainly create a very weird rendering no human on the planet has ever seen because our eyes won't let us. In sunsets we're not suppose to see everything because there is less light illuminating the overall scene.
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: HDR workflow question
« Reply #7 on: November 29, 2013, 04:29:18 pm »

It's clear some folks here would go through the trouble of using a tripod and stacking bracketed exposures even with a 14+EV DSLR instead of employing both Parametric and Point curve in ACR/LR which is far easier to do with a single exposure.

I've done quite a few single exposure sunset landscapes using a 2006 8EV DSLR recovering enough detail on both ends of the luminance spectrum even on jpegs just exposing for highlights so I'm not getting what folks are wanting to achieve with even better equipment.

Our eyes have to adapt to either the brightly lit or the shade as in the OP's linked boat dock image. Forcing an adaptive appearance in post for both dark shadows and highlights will certainly create a very weird rendering no human on the planet has ever seen because our eyes won't let us. In sunsets we're not suppose to see everything because there is less light illuminating the overall scene.

The 14.3 stop dynamic range that DXO reports for the D800e is optimistic. That is the engineering DR with a SNR of 1.0 in the darkest shadows, but that amount of noise is hardly acceptable for most purposes and the usable photographic DR is a couple of stops less. And not everyone has the D800e. Here are some real world DRs measured by Bill Claff. IMHO, HDR is still needed in special situations.

Bill

Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: HDR workflow question
« Reply #8 on: November 29, 2013, 04:56:55 pm »

The 14.3 stop dynamic range that DXO reports for the D800e is optimistic. That is the engineering DR with a SNR of 1.0 in the darkest shadows, but that amount of noise is hardly acceptable for most purposes and the usable photographic DR is a couple of stops less. And not everyone has the D800e. Here are some real world DRs measured by Bill Claff. IMHO, HDR is still needed in special situations.

Bill



That type of DR analysis doesn't tell me anything I can use in creating a high dynamic range image that translates to acceptable quality which is a subjective assessment at best.

Take pictures of two prints of an HDR processed image one showing the results using single exposure vs "Stack & Blend" processing and post them here and I'll be able to learn something I can use.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: HDR workflow question
« Reply #9 on: November 29, 2013, 05:20:45 pm »

Hi Bill,

Single exposure:


Three exposures HDR:


Best regards
Erik


That type of DR analysis doesn't tell me anything I can use in creating a high dynamic range image that translates to acceptable quality which is a subjective assessment at best.

Take pictures of two prints of an HDR processed image one showing the results using single exposure vs "Stack & Blend" processing and post them here and I'll be able to learn something I can use.
Logged
Erik Kaffehr
 

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: HDR workflow question
« Reply #10 on: November 29, 2013, 05:29:53 pm »

Are those pictures of prints, Erik? I've seen your demo before and I favor the single exposure, but that's not what I was getting at.

I was wanting a close up shot of detail from a print that shows the advantages of using "Stack & Blend" over single exposure processing. If you can't see it in a print what's the point of going through the trouble of dragging a tripod around and laboring in post trying to get all those exposures to line up and blend without kicking up artifacts.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: HDR workflow question
« Reply #11 on: November 30, 2013, 03:39:26 am »

Hi Tim,

No, they are not scanned prints, just JPEGs from my web site. Personally I think they are quite similar, except the HDR image having some artefacts.

I am not "anti-HDR", but I have not really came up with many images that really needed HDR. If I do HDR I use Photoshop to render a HDR file but use LR to tone map it.

I wanted to write an article on HDR, but failed to find some images to write about.

The image below is a HDR.

Best regards
Erik


Are those pictures of prints, Erik? I've seen your demo before and I favor the single exposure, but that's not what I was getting at.

I was wanting a close up shot of detail from a print that shows the advantages of using "Stack & Blend" over single exposure processing. If you can't see it in a print what's the point of going through the trouble of dragging a tripod around and laboring in post trying to get all those exposures to line up and blend without kicking up artifacts.
« Last Edit: November 30, 2013, 03:44:04 am by ErikKaffehr »
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: HDR workflow question (an now a single shot image)
« Reply #12 on: November 30, 2013, 04:44:16 am »

Hi again,

Here is a single shot image quickly processed in LR5, including actual pixels crop of shadow detail and highligts detail. Note snow cristalls in the sky on the highlights crop.

Best regards
Erik
Logged
Erik Kaffehr
 

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: HDR workflow question
« Reply #13 on: November 30, 2013, 09:26:26 am »

That type of DR analysis doesn't tell me anything I can use in creating a high dynamic range image that translates to acceptable quality which is a subjective assessment at best.

Perhaps that is because you do not understand quantitative analysis of noise and dynamic range. You can read up on this topic here. How much shadow noise that one can tolerate is somewhat of a personal choice, but as Emil's Figure 13 shows, a SNR of 1 is pretty useless, whereas a SNR of 8 is not bad. Each increment of SNR involves 1 stop less in DR. If you start with an engineering DR of 14 stops (SNR = 1), but require a  shadow SNR of 8:1, you lose 7 stops of DR and have 7 left. If your subject exceeds this range, you should use HDR.

Regards,

Bill
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: HDR workflow question
« Reply #14 on: November 30, 2013, 04:24:58 pm »

Hi Tim,

No, they are not scanned prints, just JPEGs from my web site. Personally I think they are quite similar, except the HDR image having some artefacts.

I am not "anti-HDR", but I have not really came up with many images that really needed HDR. If I do HDR I use Photoshop to render a HDR file but use LR to tone map it.

I wanted to write an article on HDR, but failed to find some images to write about.

The image below is a HDR.

Best regards
Erik



When you say the first image is HDR are you saying you used "Stack & Blend" multi-exposure processing for comparison to the LR5 single exposure rendering in your second post? The LR5 version really kicks butt over what I can see in foreground tree bark detail of the first. OTOH noise can be easily hidden within the texture of the dark tree bark. It's still a clean, natural looking rendering but I can't help to think that foreground tree shadow detail could be lightened up a bit more without halos and over emphasized local contrast.

With regard to landscapes I also find it difficult identifying an HDR scene that is more challenging than interior architectural scenes lit only by window light with no interior fill. Landscape's outdoor diffused light scattered about seems to bring out more similarly backlit shadow detail than low lit interior shots.

The OP's boat dock scene I would think qualifies as a more practical HDR in that the brightly lit boats present more a challenge in preserving definition/detail in their forms while maintaining brightness over recovering one flat yellow ball of the sun in a straight on sunset shot.

I found a similar sunset shot of mine to your HDR snow scene but recovered and tonemapped from a single shot 6MP jpeg with incamera contrast set to max reduction which preserved quite a bit of shadow detail and highlights. I've included a 100% crop of similar shadow detail which I thought the lens flare would've ruined at the time of capture. Still don't know if it qualifies as HDR though.
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: HDR workflow question
« Reply #15 on: November 30, 2013, 04:36:47 pm »

Perhaps that is because you do not understand quantitative analysis of noise and dynamic range. You can read up on this topic here. How much shadow noise that one can tolerate is somewhat of a personal choice, but as Emil's Figure 13 shows, a SNR of 1 is pretty useless, whereas a SNR of 8 is not bad. Each increment of SNR involves 1 stop less in DR. If you start with an engineering DR of 14 stops (SNR = 1), but require a  shadow SNR of 8:1, you lose 7 stops of DR and have 7 left. If your subject exceeds this range, you should use HDR.

Regards,

Bill

I understand what those charts and graphs are trying to do, Bill.

It's just I can't see how to use the data when I'm out photographing scenes lit under constantly changing contrast and brightness ratios and have this data tell me whether I'll get enough useable detail among all the ensuing noise in order to render a reasonably pleasant looking image containing all the dynamics captured in the scene BEFORE I trip the shutter or decide to move on to other scenes less a challenge to render in post.

Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: HDR workflow question
« Reply #16 on: November 30, 2013, 04:37:28 pm »

Hi Tim,

One scenario where I would suggest HDR to be helpful is a large dark room with some external highlight, like sum shining trough a window. Exposing for highlights would leave the dominant dark area with low exposure, making it noisy. In images like that I would suggest HDR may help.

Best regards
Erik

When you say the first image is HDR are you saying you used "Stack & Blend" multi-exposure processing for comparison to the LR5 single exposure rendering in your second post? The LR5 version really kicks butt over what I can see in foreground tree bark detail of the first. OTOH noise can be easily hidden within the texture of the dark tree bark. It's still a clean, natural looking rendering but I can't help to think that foreground tree shadow detail could be lightened up a bit more without halos and over emphasized local contrast.

With regard to landscapes I also find it difficult identifying an HDR scene that is more challenging than interior architectural scenes lit only by window light with no interior fill. Landscape's outdoor diffused light scattered about seems to bring out more similarly backlit shadow detail than low lit interior shots.

The OP's boat dock scene I would think qualifies as a more practical HDR in that the brightly lit boats present more a challenge in preserving definition/detail in their forms while maintaining brightness over recovering one flat yellow ball of the sun in a straight on sunset shot.

I found a similar sunset shot of mine to your HDR snow scene but recovered and tonemapped from a single shot 6MP jpeg with incamera contrast set to max reduction which preserved quite a bit of shadow detail and highlights. I've included a 100% crop of similar shadow detail which I thought the lens flare would've ruined at the time of capture. Still don't know if it qualifies as HDR though.
Logged
Erik Kaffehr
 

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Re: HDR workflow question
« Reply #17 on: December 08, 2013, 08:13:37 pm »

Are those pictures of prints, Erik? I've seen your demo before and I favor the single exposure, but that's not what I was getting at.

I was wanting a close up shot of detail from a print that shows the advantages of using "Stack & Blend" over single exposure processing. If you can't see it in a print what's the point of going through the trouble of dragging a tripod around and laboring in post trying to get all those exposures to line up and blend without kicking up artifacts.

The only advantage of multiexposure is noise reduction in the low light areas of the final image when preserving the highlights. If for a given scene/camera you can obtain a final image (screen or print) where noise is not a problem at the same time you preserve the highlights using a single capture, adding more shots has no advantages over single exposure. In other words, only when multiexposure is necessary, it means an advantage over single exposure.


This is a scene with 12 stops of real dynamic range, captured using 2 shots 4EV apart using an 8EV Canon 350D (that means we can collect up to 8+4=12EV of real dynamic range):






A single exposure would mean blown highlights and/or excessive noise on the floor (images left), so multiexposure using that camera (images right) was necessary.




That scene can be captured today with a Sony A7R in a single exposure. So multiexposure for the Sony wouldn't be necessary.

As long as sensor's dynamic range continue to improve, multiexposure becomes less and less necessary. I'm pretty sure many users today are feeding their Photomatix software with many more exposures than those actually needed, just because of a lack of knowledge about HDR.

Regards
« Last Edit: December 08, 2013, 08:30:06 pm by Guillermo Luijk »
Logged

gerafotografija

  • Full Member
  • ***
  • Offline Offline
  • Posts: 224
    • gerafotografija.wordpress.com
Re: HDR workflow question
« Reply #18 on: December 11, 2013, 10:58:23 am »

Hi everyone, just checking in again after noticing the number of comments this topic generated. Awesome insights on stacked HDR vs. single image file HDR processing, thanks for replying.

I had a chance to visit the same berth at the marina an hour or two earlier in the day to see if the light was better to improve on the original shot I tried to get. With the low Winter sun illuminating the foreground as well as the background, I didn't see any advantage to HDR.

Although the scene turned out to not be as interesting as I originally thought, here is a follow up image for completeness sake.

Note, I opted for a little color mapping to add some drama based on an older film style, but no stacking and only a little manual luminance level tweaking with global adjustment tools instead of specific DR enhancement. I did grab +\- 1eV exposures, but didn't need to use them (used E-M5 this time instead of the X20, so may be better DR to start with).

Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: HDR workflow question
« Reply #19 on: December 11, 2013, 02:01:07 pm »

That's a much more pleasing capture of the dynamics of that particular scene, Arėjukas.

Though it would've been more informative with regard to this subject if you attempted to make the boat in shadow and the background bathed in sunlight look convincing as well as pleasing on your original capture that started this thread.

But sometimes it's just better and far easier to go back and grab better light, right?

Logged
Pages: [1]   Go Up