The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

HDR for perfectionists :)

torger

Active member
I guess this is a commercial announcement so beat me if you like... I'm a frequent contributor to this forum and I'll stay around to answer questions etc so it's not just spam, and I do hope this is of interest to some of you.

I'm an amateur photographer that shoot landscapes with my Linhof Techno / Leaf Aptus gear. My profession is in software development though, and in the recent 18 months we've had the opportunity to realize a dream, namely develop our own first commercial photo application which we've just released in version 1.0. We're making a stealth introduction as we want to be able to provide close and personal support here in the early phase.

It's Lumariver HDR, kind of an HDR software for those that actually hate HDR :)

It can merge and tonemap, but the focus is natural look, user control and highest image quality. It's not the software you use to get that gimmicky HDR look, but something you use to solve lighting and dynamic range problems typically in landscape photography, either stand-alone or as a companion with your raw converter and/or photoshop, always with the goal of a natural time-less look, ie it should not look like "HDR". A pretty unique feature is that it can import raw (various formats) merge/tonemap and then export to raw (DNG) so you can edit that in your raw converter just as if it was a regular raw.

The problem now is to find users for it. It's no coincidence that I've put quite some effort into making the feature set good for tech cam users, so it has built-in LCC correction and such. I primarily made the software to fit the type of user I am myself, ie a perfectionist control-freak with a pixel-peeping sickness :).

I think the program will attract users that are used to having a palette of post-processing tools which the mix in various ways depending on what work they're doing. The program has an open architecture, ie import/export almost everywhere so you can use the results together with other software in various workflows, we've probably not thought of all workflows ourselves.

My favorite workflow I use myself is to shoot raw (.mos) possibly with a gradient filter on so I can make a single-shot in difficult light, shoot my LCC, then import in Lumariver HDR cancel out the grad filter and everything with LCC and apply new tonemapping using the software's algorithms, possibly with some manual fine-tuning, and then export to a DNG which I then edit in a raw converter.

Unfortunately for us Phase One / Leaf / Mamiya MFDB users Capture One kind of sucks at dealing with DNG (not the same color profiles as when importing the straight raw) so in that case a TIFF workflow can be better. With Lightroom and RawTherapee and many others dealing with DNG works very well though.

If you like to do your post-processing for your important images mostly manually in photoshop, to feel that you have that control from start to end, Lumariver HDR is intented to work as a useful tool when you need some tonemapping. I wanted myself a tool that can control light more than traditional dodge-and-burn, but still keeps the result neutral and accessible, and Lumariver HDR intends to fill that gap.

You can export the tonemapping as a luminance map which you can import and use in your layer stack in photoshop (or whatever photo editor you use). We want the results from the program to feel open, neutral and easy to understand rather than mystical, so when you look at your finished fine art image 20 years from now it doesn't have a dated look of some software algorithm made in 2013.

Concerning MF raw import support we support most digital backs, including the IQ260. Leaf Credo is an exception which uses some new .IIQ revision, but other Leaf backs work fine (.mos). If you get a problem with your digital back let us know it's often an easy fix. If you're a Capture One user for their color rendition I'd recommend a TIFF workflow though as said. As Lumariver HDR works only on the luminance channel you can do tonemapping without disturbing the color rendition.

Currently the software is Mac only, and as it's all floating point it's a bit of a resource hog so I'd recommend at least 8 gig RAM if you're going to play with IQ180 images. The trial is fully functional, only with watermarked output.

I'll be around to answer questions.
 

GlenC

Member
I'm very interested in trying out your hdr program for my architectural shots. Any chance there will be a windows version soon?
 
Looks like an interesting approach to HDR. Good luck with your endeavor. I would give it a try if there was a Windows version.
 

torger

Active member
Thanks for the kind words.

Much of the code is cross-platform already (I've actually done most algorithm development on a Linux system) so we hope to be able to provide a Windows version too at some point, we'll need some time though. We also want to spend some time to expand the feature set, improve performance, make the GUI user-friendlier, create demo videos, not forget to continue photographing, as well as being able to provide personal support to the users, so we won't be out of things to do :)
 

Ed Hurst

Well-known member
Sounds very interesting - looking forward to giving it a try. Doing this stuff manually eats up a lot of my time at present :)
 
Ok, now this seems really cool, it's too bad I haven't shot much HDR in the last year though...

I've got some questions/suggestions anyway:

1. Can you save presets? and would they be necessary?
2. How many exposures can it blend/mask at once? (before crashing? :p)
3. Is there a hybrid blending and masking system in the works, and would it conceptually make sense?
4. I know of one other program with Raw-in/out, but honesty have no idea how the Raw file is re-encoded, would it ultimately impact image quality vs. outputting a 16-bit tiff instead?
5. If you load a Tiff instead of Raw, will the software properly render the ProPhotoRGB color space?
6. Does it retain metadata after saving to DNG?
7. Does it batch?
8. Does it auto-detect LCC frames?
9. (probably duh! but...) does it auto-align shots before blending?
10. Can you make this program automatically/adaptively white-balance different regions in a shot? Or at least make it manually adjustable/exportable as an extra layer? For example, keeping the WB of an interior while adjusting the view out a window to match. Beats having to export two versions of the image and then having to blend in Photoshop. CO7 and LR5 still can't do this!

I'll try the demo out myself later, but this is what I could think of off the top of my head.
 

Ed Hurst

Well-known member
This does look very exciting!

I have a question about applying this to situations in which stitching is required. At present, I shoot a bracketed sequence of shots for each position in the stitching sequence, then create a stitch within PTGui from the files relating to one single exposure value across the scene. That is saved as a template, then applied to the sets of files that correspond to the other exposure values. By doing that, the stitching parameters are exactly the same for shots of all exposures and they perfectly line up when layered - which allows me then to use layer masks for the manual HDR stage. Very time consuming but also very precise. If the sets of files for each exposure value were separately stitched (i.e. without the 'apply template' stage to make all sets match the first stitched set), the geometry of the stitched files for each exposure value would differ (since the alignment algorithm would be working with files that have quite different levels of information within them), and layering would not work. PTGui work brilliantly with the method I use.

However, it still requires the HDR work to be done by hand using masks (because I dislike the results of any automated process I have tried); in any case, by the time this is possible, I am working with processed files and not RAWs.

I wonder if your software could ease this process somewhat. One idea would be to ensure that the merging and/or tone mapping process could somehow be set up as a template and applied to other sets of files. Imagine for one part of a scene (i.e. one set of perfectly lined up files) all files (with differing exposures) were processed through your software. Then, all of the files for the other parts of the scene could be processed in such a way that the tones match (though of course the areas so-affected would be unique in each case); the result being that a single set of files that have been processed through your software could be stitched with the values perfectly and smoothly mapping across the scene. My fear, without this capability, is that if each set of differently exposed files is separately processed through your software, that they won't match up properly tonally if stitched, giving strange banding type effects in the final scene. Maybe that fear is unjustified - it's based on the assumption that the merging / tone mapping process would be determined by the unique tonal values of the files within each part of the scene (which would of course vary across the scene) rather than to give a consistently toned output across the whole scene (or rather one that varies coherently across the scene).

The other method, of course, is to do what I do now (i.e. use PTGui to stitch the files for each exposure separately - using a template) then use your software to do the HDR stage - but of course that would be working with TIFFs (or, more likely, 'Photoshop Large' format files - which I hope your software could handle) and not RAWs; presumably this would lose some of the benefit of your software. But would that approach work?

Sorry for the long, rambling explanation of this - hope it makes sense! I suspect a lot of shooters would benefit from this capability :) Essentially I am hoping the first method would work (it would allow the use of RAWs for the HDR stage and would thus potentially be better) but perhaps it's not feasible, since it assumes applying a template to tones while still applying merging / tone mapping to some very different sets of files across a scene (which perhaps is not possible)? Failing that, would the second method work?

All the best and congrats again on a wonderful concept!

Ed
 
Last edited:

dchew

Well-known member
Seriously Anders?? I shot an interior office space last week and have been struggling with both Photomatix and Ps HDR Pro. Now I am the first to admit I know very little about the controls in either of those programs since I'm new to hdr blending. But my first try this morning with Lumariver gave me significantly better results.

You even included LCC mapping! I don't really like the flat field correction I get in LR, and I hate burning a tiff through C1 just for the LCC process. But now I can do it and save as a dng back in LR! Holy crap how did you know to create exactly the software I need?!

I'm going to do a few more tests, but you have done some fine work here! Oh, and perfect timing. :)

Dave
 

torger

Active member
Will you support the .PEF files produced by the Pentax 645D?
Yes, it's supported by the raw lib, however to be sure it work well I always want to test first. Our native raw lib is based on dcraw and with that the case is that sometimes some constans may be a bit off and need fixing. I've done fixing for the Phase One backs and some other medium format stuff, but haven't actually looked specifically at the 645D yet.

We're currently working on a v1.01 update which contains some raw fixes and improved color rendering support using DCP files by the way.
 

torger

Active member
Seriously Anders?? I shot an interior office space last week and have been struggling with both Photomatix and Ps HDR Pro. Now I am the first to admit I know very little about the controls in either of those programs since I'm new to hdr blending. But my first try this morning with Lumariver gave me significantly better results.

You even included LCC mapping! I don't really like the flat field correction I get in LR, and I hate burning a tiff through C1 just for the LCC process. But now I can do it and save as a dng back in LR! Holy crap how did you know to create exactly the software I need?!

I'm going to do a few more tests, but you have done some fine work here! Oh, and perfect timing. :)

Dave
Glad to hear! The tonemapper has received most testing for landscape images so far, so I might expect that there's gains from tuning it a bit for making it better for interior shoots, but it should work okay already.

The LCC is there because I use a tech cam and thus need it myself :). It's a more advanced feature than I thought it would be when I first started implementing it. One issue you get is that you shift clipping points. Without LCC all channels clip at their expected values but with LCC you scale values over the image and thus clip points.

The easy way to avoid problem is to cut away a little highlights by reducing clipping to the lowest level after LCC. We do it the other way around though, we reconstruct up to the new level to avoid losing any highlight detail. The drawback is that you in rare occassions can get discoloured highligts in the reconstructed areas (if there was too little information to make a good reconstruction), but then you can push that to clipping in your raw converter, and lose no more than you would have done if the LCC algorithm would just clip at the lowest level. This LCC approach allows me to do quite "aggressive" LCCs, with a graduated filter still on which I do in some landscape shots.

Finding the reference color to avoid color shifts when LCC is applied is also not too easy, with shifted lenses and possibly grad filters still on etc. So in all it was quite a large feature to implement, I hadn't done it for the first release if it wasn't because my tech camera has been the main testing platform :).
 

torger

Active member
1. Can you save presets? and would they be necessary?

You can't save presets currently, but it could be something to add. Presets would be on things like which camera profile to choose on import, possibly autopicking LCC, accepted noise levels in the merger etc.

It would probably be no presets for the tonemapper itself though, as it doesn't make sense. Our tonemapper focus only on one thing: reduce dynamic range while keeping an as natural look as possible, and then it's only so much one can do (you simple select how many stops you want to compress and mix together a few algorithms) and all the current settings are scene-dependent ie no presets can be made. It's not suitable for "creative effects" type of tonemapping, but there are other softwares that do that very well already such as Photomatix. One can of course export the merged result to OpenEXR or floating point TIFF to import into Photomatix for tonemapping there. As we have a bit more advanced color handling of raw files (especially when we get 1.01 out which I'm working on!) that can be desirable.

2. How many exposures can it blend/mask at once? (before crashing? :p)

It stores all images in RAM in floating point so it becomes a bit of a memory hog when you have lots of layers. Again as the tonemapper is not focused on creative effects we do not think you'll need more than 2 - 4 layers (with *too* much range natural results cannot be had anyway), and that should be no problem. I've actually not tried to make huge stacks to see when it breaks... if you try to load 10 images from an IQ180 I would guess something bad will happen ;)

3. Is there a hybrid blending and masking system in the works, and would it conceptually make sense?

Uhmm... I'm not 100% sure what you mean by this, but what I would like to see in the software at some point is being able to adjusts the merge masks inside the software, today you need to export/import to change them. Need to think more about it, as one would not want to make some feature that people would use Photoshop to do it instead anyway.

4. I know of one other program with Raw-in/out, but honesty have no idea how the Raw file is re-encoded, would it ultimately impact image quality vs. outputting a 16-bit tiff instead?

Raw output means that we make a "cooked" DNG, ie we take the all the floating point samples and write it to a DNG, and set all metadata as if it would have been a one-shot image from the source camera. This works well for any software that have full implementation of the DNG specification. And that would be, eeehh... Lightroom and Lumariver HDR ;). With floating point DNGs you won't lose any precision.

However other DNG converters usually generally support only 16 bit DNGs, and some only with original camera range rather than the full dynamic range. This issue is discussed at length in the manual with picture examples. In most cases you don't need to worry, especially if you export a tonemapped DNG (so you won't need to do much more shadow pushing in the target software), but if you export a merged but not tonemapped file with raw reconstruct highlights (which usually appears very dark, with up to 1-2 extra stops added on top through raw reconstruction) and the target software needs original camera dynamic range you will have a problem. Then 16 bit TIFF is better. As TIFF is gamma-encoded those 16 bits give much more range than 16 bit DNG which is linearly encoded (unless enabling our special gamma-encoded DNG hack).

5. If you load a Tiff instead of Raw, will the software properly render the ProPhotoRGB color space?

Yes. The display is fully color managed, assuming you have profiled your screen. With TIFF files we make it simple, we just tunnel the color space, ie if it's prophoto in the import it will be prophoto inside the application (converted to be displayed properly on your screen, but that's only for display, internally it's the original colorspace) and prophoto in export. Raw files if exported to TIFF will be assigned prophoto colorspace.

We'll likely add possibility to change color space on export (eg open Prophoto and save sRGB) but it's not a prioritized feature for the moment as we think most working with TIFF will have a prophoto in and prophoto out workflow.

6. Does it retain metadata after saving to DNG?

Yes.

7. Does it batch?

Not currently. It's interesting for the future though.

8. Does it auto-detect LCC frames?

No, but also a thing I'm interested in to investigate the possibilities.

9. (probably duh! but...) does it auto-align shots before blending?

If you want raw output you can't align shots, as the bayer arrays would then not line up. Otherwise you can align them, but currently you need to have Hugin installed to make the feature available as we used the command line tool align_image_stack transparently in the background for this task. They do this well already and it's tough to implement that type of algorithm so we won't prioritize to make our own algorithm for this for the moment.

I use alignment myself only when I have accidentally moved the tripod between shots, we've assumed that most users interested in a HDR software like Lumariver HDR would shoot from a tripod and thus rarely need to align shots.

10. Can you make this program automatically/adaptively white-balance different regions in a shot? Or at least make it manually adjustable/exportable as an extra layer? For example, keeping the WB of an interior while adjusting the view out a window to match. Beats having to export two versions of the image and then having to blend in Photoshop. CO7 and LR5 still can't do this!

No we can't but I know about the need and is one of the interesting features to look into. If we want to be strong on interior shoots this would be a "must have". So far we've mostly focused on landscapes and then this feature is not so much needed. (White balance setting is not so user-friendly currently, we know about this and we'll make a preview view to improve that.)
 
Top