Frustrated about HDR and dynamic range?

[Hans Kruse]

When the dynamic range to be captured for a scene is larger than the sensor can capture without clipping essential highlights and having clean enough shadows after processing, then what do we do?

We do bracketing so that the dynamic range is covered for the highlights and the shadows and then use HDR programs to merge the shots and tonne map the result. It can also be done using either Photoshop or the Photomatix 32 bit plugin for Lightroom to create a 32 bit file that can be edited (tone mapped) in Lightroom. This has been available since version 4.1. Results can be good or not so good. In the case the result is not good then what do we do? Manual blending in Photoshop.

Now manual blending can be difficult to do if the different exposures are brought into Photoshop with different exposures. There is another way that uses Lightroom and the ability of Lightroom to very closely align different exposures.

The way it can be done is as follows: One exposure that has acceptable or no clipping of highlights is selected. Edit it to your liking ignoring noise and artifacts. How much noise and artifacts depends on the dynamic range and the sensor involved (and ISO). When the editing is done check if there is another exposure exposed higher. Now select all the higher exposed files from the bracket sequence and copy the edits by shift-cmd-S and select all parameters. The edited file need to be the most selected, of course. Now align exposure by shift-alt-cmd-M. Then check each exposure and find the one with the best shadows and without being more than 3 stops higher exposed. Now open the files as layer in Photoshop (any version will do). Align the layers and put the most exposed as the top layer. Add an adjustment layer and brush the areas where the highlights are blown. Then save the file and back in Lightroom you have the blended result.

Quick and easy and does not require a lot of Photoshop skill. I never saw this method mentioned before so thought I would share it

I have found that more than 3 stops between the two shots selected for blending does not work so well for the exposure alignment. It would be possible for larger dynamic ranges to do this in two steps to achieve a 6 stop difference, but I have not tried this. Make sure the white balance is the same for the blended images. This will work for scenes with movement as long as the most exposed has the right composition of the moving parts. I have been using it for landscapes. It makes wonders for my Canon But also even the D810 which I'm using will have situations where blending is better than not due to the noise in the shadows (seldom pattern noise though as in the Canon files).

Any thoughts on this?

[AlterEgo]

I have found that more than 3 stops between the two shots selected for blending does not work so well for the exposure alignment.

Fuji sensors ( like in S5Pro ) were with half of their sensels "masked" to separate them by 4 stops...

[AlterEgo]

When the dynamic range to be captured for a scene is larger than the sensor can capture without clipping essential highlights and having clean enough shadows after processing, then what do we do?

GL (he is a member of this forum) once suggested to do blending of 2 exposures before raw converter = http://www.guillermoluijk.com/article/nonoise/index_en.htm

[PhotoEcosse]

I guess we all find the methods that suit us best.

For my purposes, using Nikon D800/D800E/D810 Raw files, I find the best way is to take exposures at 2EV intervals (with those cameras, for some obscure reason, Nikon limit me to 1EV intervals in auto-bracketing, so I just discard the intermediate values). Normally + or -  5 EV will cover the dynnamic range of even quite an extreme sunrise or sunset.

I then process in HDREfexPro2 from within Lightroom. I find the Nik program vastly superior to Photomatix as it allows me to use the control points to avoid getting an "overcooked" HDR. For my purposes, a good HDR image is one where no-one can tell that HDR has been used.

[Tim Lookingbill]

Define with a sample shot what is considered dynamic range that is larger than the sensor can record. Describing in stops is too vague.

I've never had to resort to the amount of work you describe just walking around and shooting whatever interests me using a 6 year old Pentax K100D DSLR. I have over 3000 hand held single exposure Raw shots of interiors and exteriors, some with noisy shadows that can't be seen on an inkjet print because the fineness of the grain (thanks to PV2010-2012 noise removal) is smaller than inkjet droplets.

If you're talking about professional quality, prestine looking architectural interiors with exposed windows showing outdoor detail, then I'ld understand the use of exposure blending, but then I still get decent single exposure results of shots I've taken of downtown local businesses as a hobbyist.

Could you post a before and after sample capture demonstrating how well your described technique makes it worth the trouble?  

[PeterAit]

I agree with Tim - it's really not clear how the OP's technique would be preferable to what I do: 3 exposures at -2, 0, and +2 EV and then process in HDR Expose. I get very naturally appearing results (and I truly hate the HDR "look", so I am fussy) with a minimum of effort and time.

[jferrari]

Same process as Peter for me. I'm happy with the results.    - Jim

[Hans Kruse]

Define with a sample shot what is considered dynamic range that is larger than the sensor can record. Describing in stops is too vague.

I've never had to resort to the amount of work you describe just walking around and shooting whatever interests me using a 6 year old Pentax K100D DSLR. I have over 3000 hand held single exposure Raw shots of interiors and exteriors, some with noisy shadows that can't be seen on an inkjet print because the fineness of the grain (thanks to PV2010-2012 noise removal) is smaller than inkjet droplets.

If you're talking about professional quality, prestine looking architectural interiors with exposed windows showing outdoor detail, then I'ld understand the use of exposure blending, but then I still get decent single exposure results of shots I've taken of downtown local businesses as a hobbyist.

Could you post a before and after sample capture demonstrating how well your described technique makes it worth the trouble?  

One example is in the attached photos. First one is a -1EV as imported in Lightroom. The second is edited and the 3rd is the +1EV after the adjustment of exposure and the last one is the merged one.

The whole point of this technique is that the editing is done in Lightroom as any RAW file and the mering is quick and easy since the exposures are adjusted in Lightroom before the manual merging in Photoshop.I hope it is clear

[Hans Kruse]

Define with a sample shot what is considered dynamic range that is larger than the sensor can record. Describing in stops is too vague.

I've never had to resort to the amount of work you describe just walking around and shooting whatever interests me using a 6 year old Pentax K100D DSLR. I have over 3000 hand held single exposure Raw shots of interiors and exteriors, some with noisy shadows that can't be seen on an inkjet print because the fineness of the grain (thanks to PV2010-2012 noise removal) is smaller than inkjet droplets.

If you're talking about professional quality, prestine looking architectural interiors with exposed windows showing outdoor detail, then I'ld understand the use of exposure blending, but then I still get decent single exposure results of shots I've taken of downtown local businesses as a hobbyist.

Could you post a before and after sample capture demonstrating how well your described technique makes it worth the trouble?  

I'm shooting landscapes and it is only in rather extreme situations where this is needed anyway. But when it is and I think most landscape photographers will recognize what I'm talking about, then here is presented a different technique that makes the editing as natural as you like from Lightroom and at the same time merging to another exposure easy and quick. That's the whole point of this. If you have never tried to do this then you may not appreciate it

[jferrari]

One example is in the attached photos. First one is a -1EV as imported in Lightroom. The second is edited and the 3rd is the +1EV after the adjustment of exposure and the last one is the merged one.

The whole point of this technique is that the editing is done in Lightroom as any RAW file and the mering is quick and easy since the exposures are adjusted in Lightroom before the manual merging in Photoshop.I hope it is clear

Hans, by adjusting the exposure you are defeating the purpose of shooting the brackets in the first place. Also, I use LR/Enfuse to do the blending (merging) right in LightRoom.   - Jim

[Tim Lookingbill]

I'm shooting landscapes and it is only in rather extreme situations where this is needed anyway. But when it is and I think most landscape photographers will recognize what I'm talking about, then here is presented a different technique that makes the editing as natural as you like from Lightroom and at the same time merging to another exposure easy and quick. That's the whole point of this. If you have never tried to do this then you may not appreciate it

Which of the four is strictly edited in LR as a finished image without resorting to merging bracketed shots as a comparison to show whether there are benefits gained using the extra steps?

I can hardly tell a difference between the 2nd and 4th image.

[ErikKaffehr]

Hi,

Pushing the limits of an exposure intended to keep highlights can cause noise in the shadows, which may be masked by downscaling for web size images but are very obvious at actual pixels.

Using HDR may be a good solution, but HDR has a couple of extra problems.

- Ghosting is possible if there is anything that moves.
- Tone mapping is needed

In many cases it amy be better to use the technique that Hans describes. Use an image optimised for the highlight and blend manually with one having better shadow detail.

I have run into a situation similar to what Hans describes recently and resolved it with a method pretty similar to what Hans describes.

Horses for the courses.

Best regards
Erik

I'm shooting landscapes and it is only in rather extreme situations where this is needed anyway. But when it is and I think most landscape photographers will recognize what I'm talking about, then here is presented a different technique that makes the editing as natural as you like from Lightroom and at the same time merging to another exposure easy and quick. That's the whole point of this. If you have never tried to do this then you may not appreciate it

[Bart_van_der_Wolf]

Hans, by adjusting the exposure you are defeating the purpose of shooting the brackets in the first place.

Hi Jim,

Not entirely. The 'exposed for the shadows' image(s) will have significantly lower noise than the 'exposed for the highlights' image. So what Hans is achieving is the creation of a better version of the highlights shot, that will stand subsequent post-processing much better than the single highlights shot.

However, you are correct that it is not the same as exploiting a floating point HDRI file, because Hans' version will become a low noise 15-bit/channel Photoshop file after merging the layers, while an HDRI can span a hugely larger range of unique (in linear gamma) values, which will allow more extreme tonemapping. Of course, before merging one can tonemap the shadows image separately, but that may introduce posterization and color issues (depending on the tools).

Also, I use LR/Enfuse to do the blending (merging) right in LightRoom.

Yes, that already uses some more of the HDR potential, although Lightroom can also do some useful stuff with a 32-bit/channel floating point TIFF. The size of the TIFF however, may be a discouragement to have it around longer than necessary (one can always remerge it, e.g. with the "Merge to 32-bit Plugin for Lightroom " from HDRSoft or their Photomatix application, if one keeps the original Raw exposure brackets around).

The merit of Hans' method is that it just improves the signal to noise level (esp. for the shadows) for a regular/natural looking image. The goal is similar to Guillermo Luijk's ZeroNoise method (but that is potentially more accurate). That already allows a bit more playroom, especially in the shadows, even if it isn't an HDR scene.

I personally like the level of control and tonemapping capabilities of real HDRI files, which is more flexible, but Hans' method is a nice approach as an alternative.

Cheers,
Bart

[PhotoEcosse]

Just as an additional point of reference, here is a brief explanation I prepared for my camera club last year which shows the basic idea (plus some of the negative effects if you take it too far).

http://www.premier-pages.co.uk/hdr/hdr.htm

[Hans Kruse]

Which of the four is strictly edited in LR as a finished image without resorting to merging bracketed shots as a comparison to show whether there are benefits gained using the extra steps?

I can hardly tell a difference between the 2nd and 4th image.

Number 2 was edited in LR and the editing copied to number 3 using cmd-shift-S and the exposure was adjusted using shift-alt-cmd-M. So number 2 and 3 are entirely adjusted in LR and only number 4 is a result of the manual blending in LR. The difference between my method and the typical manual blending is that the two images to be blended are as identical as possible before blending except for noise in the shadow areas and clipping in the highlight areas. The blending keeps all areas from the most exposed shot and only has areas from the least exposed shot where the highlights are clipped. Therefore the best of two worlds and with the undesired side effects of the typical automatic blending process.

[Hans Kruse]

I agree with Tim - it's really not clear how the OP's technique would be preferable to what I do: 3 exposures at -2, 0, and +2 EV and then process in HDR Expose. I get very naturally appearing results (and I truly hate the HDR "look", so I am fussy) with a minimum of effort and time.

There are a number of problems using HDR programs and also other automatic blending methods. I'm not suggesting that you change method if you are happy with the results, but what I am suggesting is that for those who found manual blending difficult and time consuming to look at my method because it is very quick and effortless to do.

[Hans Kruse]

Hi Jim,

Not entirely. The 'exposed for the shadows' image(s) will have significantly lower noise than the 'exposed for the highlights' image. So what Hans is achieving is the creation of a better version of the highlights shot, that will stand subsequent post-processing much better than the single highlights shot.

However, you are correct that it is not the same as exploiting a floating point HDRI file, because Hans' version will become a low noise 15-bit/channel Photoshop file after merging the layers, while an HDRI can span a hugely larger range of unique (in linear gamma) values, which will allow more extreme tonemapping. Of course, before merging one can tonemap the shadows image separately, but that may introduce posterization and color issues (depending on the tools).

Yes, that already uses some more of the HDR potential, although Lightroom can also do some useful stuff with a 32-bit/channel floating point TIFF. The size of the TIFF however, may be a discouragement to have it around longer than necessary (one can always remerge it, e.g. with the "Merge to 32-bit Plugin for Lightroom " from HDRSoft or their Photomatix application, if one keeps the original Raw exposure brackets around).

The merit of Hans' method is that it just improves the signal to noise level (esp. for the shadows) for a regular/natural looking image. The goal is similar to Guillermo Luijk's ZeroNoise method (but that is potentially more accurate). That already allows a bit more playroom, especially in the shadows, even if it isn't an HDR scene.

I personally like the level of control and tonemapping capabilities of real HDRI files, which is more flexible, but Hans' method is a nice approach as an alternative.

Cheers,
Bart

I just got the idea of the approach I described a few days ago. I have shunned manual merging for a while and have used the 32 bit blending and tone mapping in LR for a while when needed. But I often were not quite happy with the results and often preferred a straight edit of the best compromise of the bracketed sequence. This meant that I had to accept noise in the shadows and sometimes a bit more clipping of highlights than I really liked. Now with this approach I can blend two exposures easily 3 stops apart and that solves almost all the cases I have come across. That's for shots I have in my archives using the Canon 1Ds III and 5D III. With the D800E and D810 I can avoid most of these cases however I have seen noise that I'd like to avoid and in some cases even some banding. So even with these cameras there are cases where this approach is useful.

As I mentioned in another response, I'm not going to preach this is a solution to all problems My suggestion is rather: Try it out and see for yourself when it makes sense to use and when it gives better results.

I have now checked a number of my earlier blendings and I found almost always the manual blending gave the better result. This I believe is mainly due to imperfections in the algorithms used in the automatic blending approaches. I have used the 32 bit Photomatix plugin for Lightroom, LR/Enfuse and Photoshop HDR Pro 32 bit. I have also tried a number of stand alone HDR programs and rejected all of them. I have not systematically documented all of these findings, which could have been useful, of course. However my approach is to suggest other ways and techniques and people can try for themselves to see what works for them. Depending on style and taste someone else can easily come to a different result than I do.

[Hans Kruse]

Hi,

Pushing the limits of an exposure intended to keep highlights can cause noise in the shadows, which may be masked by downscaling for web size images but are very obvious at actual pixels.

Using HDR may be a good solution, but HDR has a couple of extra problems.

- Ghosting is possible if there is anything that moves.
- Tone mapping is needed

In many cases it amy be better to use the technique that Hans describes. Use an image optimised for the highlight and blend manually with one having better shadow detail.

I have run into a situation similar to what Hans describes recently and resolved it with a method pretty similar to what Hans describes.

Horses for the courses.

Best regards
Erik

Can you describe how your blending method was different than what I suggest? I'm interested in experience in other blending methods in Photoshop. What did you prefer in terms of final result, manual blending and automatic blending? As you know I have been preaching up to now to avoid HDR approaches of any kind merging bracketed shots if possible. However with this new found approach (for me I have changed my opinion.

[kers]

Hans, first - thank you for starting this thread...

as i understand :
you used 2 photographes in your sunset example;  say 2 stops different in exposure. A= Light and B= dark
The lightest one A you have developed in two different ways - only adjusting the  exposure in LR. So you have now two versions of Photo A
As a result you end up with 3 images and you mix them- as three layers-  together in photoshop.

( correct me if i am wrong)
if i am correct; the conclusion is that LR or ACR at the moment is unable to get one perfect version of photo A.
It is something i also conclude from my own experience with ACR.
it looks like that Abobe did not see coming the enormous dynamic range of some sensors- maybe in a next version it will do better.

Your method  is about how i deal with enormous dynamic range problems.. I am good with photoshop and find HDR programs usually not doing so well in getting a natural looking photo.

btw your sunset mix is very well done.

[Bart_van_der_Wolf]

Using HDR may be a good solution, but HDR has a couple of extra problems.

- Ghosting is possible if there is anything that moves.
- Tone mapping is needed

In many cases it amy be better to use the technique that Hans describes. Use an image optimised for the highlight and blend manually with one having better shadow detail.

Just to add, significant tonemapping is mostly required to fit the dynamic range of the scene in the more limited dynamic range of or output-modalities. This already applies to normal scenery which easily surpasses what e.g. printed output can handle. Real HDRI obviously poses an even bigger challenge.

In addition, shadows are more noisy by default, due to the Poisson distribution of exposure by Photons, and made worse by sensor limitations and demosaicing. It usually helps to improve technical shadow quality (unless one down-samples a large amount).

The implementation that Hans described, will help to improve overall image quality if ghosting artifacts can be managed. Other methods would involve the averaging of multiple (identical) exposures that are chosen to preserve highlights. Averaging will not only improve shadow detail, but also highlight detail, although it would require to average some 8 exposures to achieve similar shadow quality as a +3 stop exposure bracketed shot can add to the overall quality.

Cheers,
Bart