Skip to main content

HDR Artefacts

Comments

16 comments

  • gb

    Is there any movement of image elements between shots?
    I find C1 gives pretty poor results in those areas of movement.

    0
  • Charles O'Hara

    Ineed, Capture One is pretty poor at managing movement from frame to frame when merging to HDR. It's really hit or miss and you can't know for sure before trying the merge. I noticed that it struggles to handle very noisy images in the stack (say, a very dark exposure that is protecting the highlights but that has otherwise deep, noisy shadows) - it will often interpret the noise as some kind of detail and integrate it into the merge.

    Ideally, the algorithm should check the EV values of the exposures and prioritize the ones with the highest SNR (longest exposures at lowest ISO, usually), but it doesn't for some reason, so you may get blotchy images with patches of noise throughout, stepping in gradients, or just a loss of detail where there should be some (according to your best exposure SNR wise). I've had much better results by first passing my raws through DXO PureRaw and then merging the resulting DNGs in Capture One - Capture One then gives acceptable or good results most of the time.

    0
  • Mark Farrington

    Yes, I've found the same with going through Topaz Photo AI.

    I find CO's HDR merge absolutely fine for most casual use (eg social media, personal blog posts), but for critical uses (camera club competitions, printing) I usually take the time to go via Nik HDR eFex.

    0
  • Daniele

    Thank you all for your replies.
    These are indoor photos taken with a tripod at 160 ISO, but the artefacts are in a palm tree outside the window. There was no wind that day, but the leaves might have moved slightly anyway.
    I'll post a picture of the problem as soon as I get home.
    Thanks

    0
  • Daniele

    0
  • Charles O'Hara

    @Mark: I use the built-in HDRs from CO for professional architecture photography and so far nothing compares to the latitude this provides, because the resulting linear DNG behaves just like a raw file, and by demosaicing the raws in DxO PureRAW I get to keep the DNG workflow all the way, which means full dynamic range and color depth for editing in Capture One, with added lens correction and AI noise reduction at the source. All other HDR software that I know of either produce 32 bits DNGs incompatible with CO, or just output TIFF files which are vastly inferior to linear DNGs. This is why I stick with CO, even if the merging itself could use a lof of improvements: when it works, it's the best for a CO workflow IMO.

    @Daniele: Those dark spots are classic merging "mistakes" resulting from leaf movement between your shots. One possible workaround is to merge fewer shots of the bracketed sequence at the time, say merge +1EV and 0EV together, then merge the result with the -1EV (or whatever your brackets are). I noticed CO is better at merging fewer images together than a big bunch, so it might be worth a try. If that doesn't help, you can either retouch those spots manually (shouldn't be hard with the new Photoshop tools), or just export one of the raws used in the HDR and layer it and mask it out appropriately over the rendered DNG in Photoshop in order to cover the artifacts with the source image. I use all 3 solutions myself when dealing with problematic merges.

    0
  • Mark Farrington

    @Charles Yes, good points, and I agree about the much greater flexibility of DNGs when significant post-processing is required, although I find Nik much better at merging more than 3 files without artefacts and at coping with movement. (Or, at least I did when I last tried merging more than 3 in CO, which was a while ago: do you have recent experience on that?)

    There is scope for CO to develop usefully further in this area since it’s much quicker and less hassle to be able to stay within the raw converter.

    0
  • Charles O'Hara

    @Mark, I rarely merge more than 3 or 4 exposures, even if I shot more (as much as 7 at times), because it's rarely needed to get a good file with a very large DR. I noticed that CO is better at merging longer (better SNR) exposures than lower SNR ones, even if they go through the NR DxO funnel before, so I usually pick the best ETTR lower exposure, i.e. the one that barely has the ability to recover highlights to a satisfactory degree without too much blown areas (no need to see the inside of DELs or the texture of the sun, you know), then I pick 2 or 3 of the brightest exposures (sometimes more is required to get better gradients) and hope for the best.

    If stuff move too much and I get artifacts, usually I just keep a DNG straight out of PureRAW on the side, I edit the CO HDR, then copy all of its adjustments (including layers) on the DxO DNG, normalize exposure and WB between the two and then I render both files and use the single DNG to mask out the artifacts on my HDR in Photoshop. The files should align perfectly at the single pixel leve, since the lens corrections were applied at the demosaicing step in DxO (you might have problems otherwise with that method if using CO lens profiles, because often if crops a little (without the possibility of recovery) when merging raws to HDR, and then the single raw file will be larger than the HDR, making aligning more difficult.

    But yeah, that's a complicated workflow, I know. CO should just be better with HDR merging, since software like Lightroom makes it such a breeze. Don't get me started on HDR panoramas in Capture One... This just adds another layer of complexity, whereas in LR it's one click for HDR panoramas, with good ghosting removal to boot. Thank God the rest of the tools are just better in CO.

    1
  • Mark Farrington

    😉

    Yes, for sure.

    Many thanks for that, very helpful. For the early part of this year I’ve shot a lot, but almost no critical work. however, I’m about to start a large project where dynamic range will be considerable and your experience is a good lead for me.

    0
  • Daniele

    Thank you all for the valuable information, it is always helpful to compare with more experienced users. 
    The problem occurred on interior photos, where the client wanted the exteriors to be seen as well and the only way was to use HDR -2/0/+2.
    Trying with LR, the artifacts were almost absent so I was wondering if CO should be configured differently or just not at the same level in this specific case.

    Thanks again for helping me out!
    Have a great evening

    0
  • Albrecht Voss

    I have had the same problem since the introduction of HDR. The pictures are close to unusable, although I capture the brackets on a tripod and base ISO and don't merge more than 3 exposures (0,-2,+2). This leaves me very disappointed, as HDR hast been of the biggest updates within years for C1 and now I can't use it. For me as a fulltime architectural photographer this is a very important topic and I can't see why it is not on C1 priority list. I have multiple shots on every shoot, where I have to merge HDR-Panos and it just does not working reliable. Very disappointing!

    0
  • Charles O'Hara

    Albrecht Voss Have you tried 5 exposures 1 stop apart instead of 3 esposures of +2/-2 EV? I've had much better results with the CO HDR algorithm when most of the merged photos are closer on the exposure range, especially when it comes to the gradients in the highlights and the handling of movement between frames. At least, that's been my experience. The worst case scenarios seem to occur with very long exposures and large gaps between exposures, say at dusk with 2s, 8s and 30s exposures - some manual blending on top of the HDR or even just manual blending is required most of the time for cases such as these.

    I'm also a full time architectural photographer and I've managed to make the HDR work acceptably most of the time, but every merge requires quite a bit of trial and error of which exposures to select and by running the raws through DxO PureRAW 3 (with DeepPRIME XD) first, in order to not let CO mistake noise as detail. 

    0
  • Albrecht Voss

    Charles O'Hara thank you for your comment. I will try this for the next shoot. Using DxO in addition would be too time consuming for me. If it doesn't work in C1, I use Lightroom, as this handles HDRs and Panos pretty reliably. 

    0
  • Charles O'Hara

    Yeah, DxO Pureraw is slow and it may not fit every workflow, but since it's a preprocessor based on an automated process I managed to make it work in the background as I do other tasks (emails, etc) and when it's done, the linear DNGs it produces work just as raw files in CO, you can select the color profile, the tone curve, etc, just as you would a raw file. Only the lens distorsion, the noise reduction (world class) and the sharpening are baked in - on top of the demosaicing process of course. It works with Lightroom too, btw, since DNGs are software agnostic.

    How do you manage different (and somehow competing) workflows in your work? Surely there are things that are more easily achieved in CO vs Lightroom and vice versa, and I guess the camera color profiles are not 100 % equivalent, nor are the tools? I'm curious, because the beginning of my career I flirted with the idea of using different editing softwares for specific purposes, but I quickly decided to focus on Capture One exclusively, with the additional help from Photoshop on top of it for the things it can't do well or at all. This gave me a uniform rendering pipeline and a unified set of editing tools for all situations and in the end, a more uniform body of work.

    0
  • Albrecht Voss

    Hey Charles, I try to use as few software applications as possible for the same reason and stick to one workflow for each project. It would be great to use C1 for everything, but it just has no all features that I need. For my medium format work with Hasselblad, I have to use Phocus as the files are not accepted by C1. If GPS meta data and location tracking is important I have to use Lightroom. Who knows, maybe there will be a time when all software will be equal in tools and you just decide by layout and workflow....

    0
  • Charles O'Hara

    Albrecht Voss I see! I doubt the tools will eventually be the "same" on every software, which is why I stick to C1, despite its flaws in some areas (which I try to alleviate to some extent with raw preprocessing).

    In your case, converting the Hassleblad raws to linear DNGs first would allow you to edit them in C1, but if you prefer to use three different editing softwares each with different tools and different handling of color, I guess it's fine by me, to each his own :)

    0

Please sign in to leave a comment.