Wikipedia:Featured picture candidates/Gravel pit

Gravel pit in Denmark edit

Voting period is over. Please don't add any new votes. Voting period ends on 26 Nov 2014 at 15:39:39 (UTC)

Original – An evening panorama of Kongensbro gravel pit near Ans in Denmark

Edit 1 by Diliff - Reprocessed. All adjustments made in Lightroom prior to exporting and stitching in PTGui

Reason: Visually attractive illustration of a gravel pit with many identifiable sub-systems (see annotations).
Articles in which this image appears: Gravel pit
FP category for this image: Wikipedia:Featured pictures/Places/Panorama
Creator: Slaunger

~~Support as nominator~~ – Slaunger (talk) 15:39, 16 November 2014 (UTC)[reply]
- Support both versions, prefer edit1 -- Slaunger (talk) 19:46, 18 November 2014 (UTC)[reply]
Comment. Not convinced that the sky is realistic looking, particularly on the left side. To my eye, it looks like the tonality is too compressed, a typical issue when reducing highlights too far. Looking at the file description, it's helpful to see your processing workflow, and I'm not sure that it was the best way to do it. I wouldn't rate PTGui's ability to tone map to be honest. I use PTGui to create an HDR file but I always re-import the HDR file back into Lightroom to tone map. Lightroom is better and gives you much more control over things. Given that this isn't true HDR, I don't see any need to use PTGui for anything other than stitching. All tone mapping processing can and should be done in Lightroom prior to stitching. Other than that though, the composition has merit and I don't want to oppose when an improvement in processing could be achieved. Ðiliff «» (Talk) 19:17, 16 November 2014 (UTC)[reply]
- Thanks, Diliff for your always very insightful reviews and I acknowledge the critique of workflow. Just some details though: The sources images looked very dull at default import settings, and I was very pessimistic regarding the potential outcome. The sky was much more exposed than the ground, which was mostly in shadow, and there was very little structure in the sky and hardly any details in the vegetation, and colors of the gravel were dull. I therefore elaborately and dramatically pre-processed the images prior to export in 16-bit tiff to PTGui Pro to basically spread out the histogram instead of having it piled up at the ends. (Highlights: -100(!), shadows: +40, contrast: -30 (yes, negative, to get midtone tails into the middle), increased clarity +30 and vibrance +15. In addition, I applied a graduated filter in the sky to further increase the contrast here and bring the exposure down). Not an optimal place for the graduated filer as it were handheld photos. Thus, the alignment from picture to picture was not perfect, but I prioritized this to feed in the most optimal dynamic range in the stitching process. I had exhausted my adjustment possibilities in the source images in Lightroom, which is why I then used the pseudo-HDR tone-mapping in PTGui to further bring out some details. Yes, I could maybe have skipped this in PTGui and have waited with the final stitched tiff and do another cirle in Lightroom. I did some final minor tweaks in Lightroom, such as quite aggressively denoise the sky with an adjustment brush leaving out the, I think, seven flock of migrating birds and also downsample some to get rid of some pretty severe noise. It was a long, iterative process of re-exporting the source images, stitch, twaek the PTGui HDR, tweak further in Lightrrom again, and I am very hesitant to give it another go. So, yes, maybe the sky to the left has too compressed colors due to dramatic lowering of highlights if viewed at full resolution. My main priority has been to give it wow and impact, not to make a color-calibrated representation of the photos that scattered on my sensor that evening:-). And if reviewers find I have been bending reality too much, I certainly respect if this leads to an oppose. -- Slaunger (talk) 20:28, 17 November 2014 (UTC)[reply]
  - I'm not sure what the problem is with the sky being much more exposed than the foreground. This is normal. ;-) I think you could have rescued the highlights in the sky without making them as murky as you did. I appreciate that you don't want to start the processing again from scratch. I've been there, done that, and it's not fun. Especially when you fix one problem and accidentally introduce another! It's images like these that I'd love to get a hold of the original RAW files and see what I can weave out of it. Perhaps I'd end up with an image no better than this one anyway, and I also appreciate that photographers can be a bit possessive with their images (not to mention their artistic decisions). Ðiliff «» (Talk) 23:11, 17 November 2014 (UTC)[reply]
    - Diliff: I do not mind giving access to the raws, if you or someone else wants to give it a try;-). Let me look into that tomorrow. Is there a recommended way to share raws for Commons files? I seem to recall Dcoetzee setup an archive server at some point... --Slaunger (talk) 23:35, 17 November 2014 (UTC)[reply]
      - I usually just use Dropbox or something. It's easy enough to just zip them up and send a Dropbox link (by email if you prefer). Ðiliff «» (Talk) 23:54, 17 November 2014 (UTC)[reply]
        Diliff Dropbox it is then! You or any other Commons user may give it a try. Please read the text file for conditions of use and some practical information. It is actually dng files (Digital Negative) (raw + metadata of my pre-processing edits in one file, instead of separate raw and xml sidecar files), so remember to reset the Develop settings if you want to start out with a clean sheet of paper. Looking forward to hearing what you can get out of it. -- Slaunger (talk) 17:18, 18 November 2014 (UTC)[reply]
        OK, I've had a go and although it was a challenging image to work with (you're right, the final image is significantly different to the 'raw product'), I think I've managed to improve the tonality while remaining reasonably faithful to your original artistic intention. There are differences though, and I'm not sure whether you will consider them improvements. Significantly, the sky is crisper and lighter and with a cooler white balance (I felt this was looked more natural but as I wasn't there, I can only guess). The forest in the background is less contrasty and greener. The gravel pit itself is fairly similar, although less saturated and again with a cooler white balance. The gravel seemed very pinkish in the original. The foreground bushes are a bit darker, greener and more contrasty. I didn't intentionally make them look greener, but I felt the original was a little washed out and the colours of the bushes suffered as a result. Finally, I was able to (with a bit of content aware fill) recover a more of the sky which reduced the aspect ratio a little (which is why it looks less wide). I usually try to maximise the height of panoramas when possible because a very wide panorama can be awkward to view and use. Slaunger, let me know if there's anything you're not happy with. I still have everything set up and ready to make minor adjustments if you think it could be improved. Ðiliff «» (Talk) 19:19, 18 November 2014 (UTC)[reply]
        Diliff: WOW, just WOW! This is a significant improvement, and I am impressed, considering the not so promising source images. The color is better on the gravel. The content aware fill is an improvement to the aspect ratio. The colors and texture of the fore- and background vegetation is improved. The sky is good. You have even managed to get out, I think, most of the many flocks of migrating birds in the sky. Your sky is more realistic than mine in the left side, although I think my more yellow right end gives a more cool gradient in the sky. Thanks, Thanks, THANKS! -- Slaunger (talk) 19:43, 18 November 2014 (UTC)[reply]
        Glad you liked the processing! I wasn't sure if it was going to be to your tastes. I realised that I didn't mask the blurry frame as completely as you did, so I might upload a new version over the top of it with the blurriness minimised as much as possible. I could easily adjust the white balance of the sky on the right side of the frame if you prefer, but it would also have the effect of making the blue sky a bit less accurate. I'm not sure what this means for the image in terms of your featured/quality pictures on the original image though. If you're happy with my version then it would make sense to overwrite the original, but the evaluation of the FP/QI was done on the older version. How do you feel about that? Ðiliff «» (Talk) 08:28, 19 November 2014 (UTC)[reply]
        If Slaunger agrees it is an improvement then I don't personally see any merit in retaining two separate files on Commons, nor for Commons FP to refer to the weaker one (should everyone agree on that). You could post on the talk page of Commons FP to see if this has community approval without the hassle of a full delist/replace. -- Colin°^Talk 08:40, 19 November 2014 (UTC)[reply]
        True. I suppose it's up to Slaunger as for how he wants to handle it. It's his image and his FP/QI, I just made some adjustments. Also, I've just uploaded a new version of the edit 1 image over the top of it. It contains a number of improvements (including a warmer sky on right side as per Slaunger's suggestion). Ðiliff «» (Talk) 09:32, 19 November 2014 (UTC)[reply]
        Personally, I would not mind simply overwriting my original nomination with Diliffs edit. It is potentially controversial though according to Commons policy on overwriting files with FP status. I have therefore requested permission over at COM:FPC to get a few nods. Else, I do not think it is a big problem to keep both either, as the preferences for featured status varies a bit between EN:FPC (faithful-oriented) and COM:FPC (wow-oriented). It is seen before. -- Slaunger (talk) 17:10, 19 November 2014 (UTC)[reply]
        
        Diliff and Colin After thinking about it and also seeing the reponse on Commons, I do not think you should overwrite my original. Just keep them separate. It will help maintain transparency in seeing the sequence of events with my original followed by Diliffs edit being linked together as 'other versions'. Also, I will follow the replace and delist process on Commons to see if there is consensus there to switch over to the edit. From the very positive response on the edit that seems likely for the time being. -- Slaunger (talk) 21:50, 19 November 2014 (UTC)[reply]
Support I agree with Diliff that the sky processing isn't ideal (per my Commons review) but not troubling enough for me to oppose a scene with such wow. Btw, Diliff did you mean to write "All tone mapping processing can and should be done in Lightroom prior to stitching." (my emphasis). I worry that since the global adjustment sliders are in fact content-aware tonemapping controls, there is a risk of tonal variation among segments even if settings are sychronised -- though whether this happens in practice I don't know. Not sure there is a huge difference to delaying tonemapping till afterwards if one is dealing with 16-bit tiffs as intermediate files. Can't comment on PtGui vs Lightroom's tonemapping abilities, though. -- Colin°^Talk 13:11, 17 November 2014 (UTC)[reply]

A technical discussion about the fidelity and dynamic range in raws vs 16-bit tiff files not strictly related to the nomination

- Colin, I suppose I should clarify, I meant specifically in situations where you're not working with a 32 bit HDR file, the processing should be done in Lightroom prior to stitching because all processing should be done in a file that contains the most fidelity. Because Slaunger exported the files from Lightroom to a 16 bit TIFF, he lost fidelity (dynamic range) prior to tone mapping, which I suspect has caused the problems. Tone mapping from a low fidelity image is usually going to be inferior to higher fidelity images (eg a single RAW file, or 32 bit TIFF generated from multiple exposures). In situations where you want to do true HDR tone mapping from multiple exposures, it's best to stitch first and output the panorama as a 32 bit file, then do the tone mapping afterwards because there is no loss of fidelity prior to tone mapping. Lightroom might be content-aware but I've never had any significant differences between component images in a panorama when using that workflow on a RAW file in Lightroom. I've only had issues when tone mapping individual segments of a panorama in Photomatix, but since Slaunger was not using Photomatix, I didn't think to mention the exception. Ðiliff «» (Talk) 13:31, 17 November 2014 (UTC)[reply]
  - I'm not sure I'd class a 16-bit tiff as "low fidelity". It should have sufficient dynamic range to hold the output from a ~12-bit raw file. The major difference I believe is that the tiff/jpg has a gamma applied to award more bits to the tonal range the eye can see (mid tones) rather than shades of blinding white, and it has a standard colourspace applied. But those 16-bits are loads compared to the 8-bit JPG we're all viewing -- if 16-bits was considered greatly inferior to raw for post-processing purposes, then Lightroom would have an "Edit in Photoshop - export as 32-bit tiff" option. It could in principle be weaker than the source but whether the difference is visible I don't know. I'm more tempted to pin the blame on PTGui's tone mapping being not as good as Lightroom at sensitivity rescuing highlights. Or, cough, weaknesses in Canon sensor dynamic range :-) -- Colin°^Talk 20:50, 17 November 2014 (UTC)[reply]
    - I've just noticed Slaunger's reply above. That's a strange mix of settings/steps for sure. I wonder if you've adjusted the histogram rather than the image, if you follow what I'm suggesting -- so tempting since the histogram is just above the sliders. But if a compressed tonal range is what you artistically aimed for then that's your choice. -- Colin°^Talk 21:05, 17 November 2014 (UTC)[reply]
      - Colin: No, I targeted to bring out information and wow in the source images and that happened in this case to relate quite strongly with spreading out the histogram - it is still far from flat. -- Slaunger (talk) 21:41, 17 November 2014 (UTC)[reply]
    - A 16 bit TIFF is still potentially lower in fidelity than a 12 bit RAW file because all the processing settings are 'locked in' (white balance, colour space, contrast, etc). Also, as I understand it, exporting a RAW file to 16 bit TIFF crops the dynamic range, so what you see as black in Lightroom equals a value of 0 in the TIFF, and white equals a value of 65536. Unlike a RAW file, there is no extra dynamic range 'hidden' beyond what is viewable on screen (assuming the screen is capable of viewing the full range, of course), so unless you convert to 16 bit TIFF with a gamma of 1 (truly linear), you will either be compressing the dynamic range or cropping it. Realistically, this means that a 'regular looking' processed 16 bit TIFF file does not contain as much actual information as the RAW file that it came from, even if 16 bits per colour channel should in theory be capable of holding all the luminosity data of the RAW file. Anyway, we're getting pretty academic at this point. My point still stands though, processing from RAW is better than 16 bit TIFF. :-) Ðiliff «» (Talk) 23:11, 17 November 2014 (UTC)[reply]
      - On reading a bit more, it appears the tone-curve applied (the gamma), is the biggest factor limiting the ability to recover highlights from a 16-bit TIFF, as it moves so much precious information away from the white end. I have stitched image with challenging DR that I plan to rework anyway so I will this time apply my global adjustments prior to export to see if the highlights and shadows are handled better. -- Colin°^Talk 08:26, 18 November 2014 (UTC)[reply]
        Having also thought about it a bit more, I've reached the conclusion that losing highlights/shadows is unavoidable with a 16 bit TIFF because unlike RAW, it's a file format intended to be directly viewable on screen. It may have the bit depth to in theory store the data of a wide dynamic range but (assuming you did store the image luminosity linearly without gamma applied) it would not contain the requisite instructions for an image viewer to convert it into a gamma and luminosity range that is viewable on screen. An image viewer/editor assumes the gamma has already been applied, so in a sense, it is a high fidelity but low dynamic range viewable format - not an archival format like a RAW file. I suppose it therefore was wrong to call it low fidelity when it is actually the dynamic range that limits the format, not the fidelity. In practice, the fidelity is lower, but not because of its bit depth. :-) Ðiliff «» (Talk) 10:27, 18 November 2014 (UTC)[reply]
        The question isn't whether a 16-bit TIFF can represent wide dynamic range or not since DR is independent of bits. One could have a 4-bit format with huge DR but terrible precision. An 8-bit JPG can represent more than 8-stops of light, because of the gamma. The question is whether the format stores that range with fidelity. There's nothing built into TIFF that knows the brightness of white or the darkness of black -- that depends on the settings on your monitor and brightness of the LCD backlight. I think, for the purpose of highlight recovery, it is the gamma that reduces the fidelity of a TIFF at the white end. But also, the magic that Lightroom can do to recover highlights provided not all the channels are blown -- that may well rely on Bayer pattern demosaicing -- once you transform that pattern to a solid pixel and apply a colour temperature and smaller colourspace, you've lost the ability to do that magic. But I'm guessing here. -- Colin°^Talk 11:20, 18 November 2014 (UTC)[reply]
        The question absolutely is whether a 16 bit TIFF can represent a wide dynamic range though. Not theoretically, but in practice. I have already accepted that a 16 bit TIFF has the same fidelity as a RAW file, it's the dynamic range that suffers. That's the reason why I've been suggesting all along not to export from Lightroom prior to tone mapping. In practice, it cannot have a high dynamic range, because it's got a fixed gamma and is tied to fixed luminosity values. A RAW file does not, and is not limited in the same way. I'm not talking about the blackness of black and the whiteness of white, I'm saying the maximum value of luminosity in a 16 bit TIFF is 65536. This value does usually correspond roughly with the whitest white that your monitor can display when calibrated properly (which is 255 given that your video card downsamples it to 8 bit when sending it via DVI/Displayport anyway). If it didn't, your images would have inappropriate areas of overexposure. I think the confusion here is that while RAW and 16 bit TIFFs can in theory contain any dynamic range that you care to feed it, 16 bit TIFFs are not viewed or processed in that way and are mapped to the luminosity of the display via gamma correction. The dynamic range is clipped when a RAW file is converted to 16 bit TIFF because it is given a tone curve and gamma that is standardised. It's not possible to have a 16 bit TIFF that contains as much dynamic range as a RAW file while still looking 'normal', except of course via tone mapping which of course creatively compresses it. Ðiliff «» (Talk) 12:32, 18 November 2014 (UTC)[reply]
        See Gamma correction. Gamma is a curve rather than a straight line but it still goes from 0 to 1. The luminosity of 1 is still "max" as it was on the sensor as it is in your JPG and on the whitest pixel the monitor can display or the pure white of your paper. TIFF knows nothing about lumens so there's nothing fixed. The reason I say the curve is the problem is that with a linear encoding we may have 500 values between "extremely white" to "blindingly white". Once you apply a curve and generate an 8-bit JPG, all of them get compressed within the values 254-255, say, and it looks awful. On a 16-bit TIFF it might be 65520 to 65535. So those 15 integers have to represent all tonal variation in your brightest clouds. This still looks bad on your 8-bit display so you decide to reduce the highlights. Once you crank the slider down, the curve is adjusted so the pixel that was at 65520 is now 65000 and the 15 shades above are represented by 535 shades but with huge gaps and with a stubborn blob in the middle of the cloud that remains 65535. This isn't really "clipping" as nothing was tipped "over the threshold" but simply that lots of useful information was compressed into a small number of integer data values and this loss can't be recovered. As the Gamma correction article says, if the TIFF was floating point, this wouldn't be a problem as FP represents numbers in a way that is independent of their magnitude. -- Colin°^Talk 13:33, 18 November 2014 (UTC)[reply]

Support a very nice and interesting photo and per Diliff (also an interesting info from Diliff!!!). --Alchemist-hp (talk) 19:57, 17 November 2014 (UTC)[reply]
- Support Edit1 - Diliff-version. --Alchemist-hp (talk) 23:24, 18 November 2014 (UTC)[reply]
~~Tentative support~~ seeing what Diliff can get our of this. — Crisco 1492 (talk) 01:17, 18 November 2014 (UTC)[reply]
- Support edit1: I have to agree with EuroCar; the edit makes it look more real. — Crisco 1492 (talk) 00:12, 19 November 2014 (UTC)[reply]
Support either versions - nice! Adds EV in terms on illustrations. Diliff's edit makes the image feel more real. ///EuroCarGT 01:37, 18 November 2014 (UTC)[reply]
Support any, prefer Diliff Hafspajen (talk) 18:07, 18 November 2014 (UTC)[reply]
Support edit 1. Forgot to actually support this nom. :-) Ðiliff «» (Talk) 23:00, 20 November 2014 (UTC)[reply]

Promoted File:Kongensbro gravel pit 2014-09-17 Diliff Reprocess.jpg --Armbrust ^{The Homunculus} 15:40, 26 November 2014 (UTC)[reply]

There is a clear consensus that Edit 1 should be promoted. Armbrust ^{The Homunculus} 15:40, 26 November 2014 (UTC)[reply]