a 1 EV bracketed, 7 stills HDR, which I could not manage to satisfactorily stack with gin and hugs. In Photoflow I simply applied (GMIC’s) Provia LUT. I ended up exporting this jpeg from PS as I also needed a smaller copy and for downscaling I always use Sverlov’s C3C
Gentlemen, a good day to y’all
PD
the tittle it’s a bit dramatic, though the kitten’s quite traumatised, the eyes full of purulence, his/her body skeletal and plenty of bugs… it will get better… or die with his/her belly full =)
Very cute! Hopefully the little guy can pull through!
What difficulties did you have in stacking the images? What method did you use?
@paperdigits I batched the images with dcraw -v -4 -T and then actually followed your comprehensive footsteps. The output was a bit misaligned, but my HDR’s exrs out of hugin have always a a kind of “posterisation”… anyways it’s too much hassle for my taste; ddd b700 — that was the kitten on the keyboard, he squeaks like a toy 
Ah yes, very good. If you have a bit of movement in your stacks, hugin’s masks work really well.
Hello chroma_ghost,
THANKS a lot indeed for your post : I was not aware about: Sverlov’s c3c plugin.
It looks really powerful ! 
Unfortunately, It looks like it only works with Photoshop (which I don’t own…). No stand alone version in short to try it 
With Gimp 2.9.X (Gegl) there is a “new” option to downscale the images : No-halo and Lo-halo samplers [1] and it would be interesting to compare the two softwares (Gimp vs C3C plugin with Photoshop)…
[1] http://www.gimpusers.com/tutorials/whats-new-in-gimp-2-10
Hey Silvio, yes it is… I’ve experimented with quite a bunch of downscaling algos from OSS to expensive tools and personally I always find Sverlov’s the best, in the sense of balance, specially regarding granny images (I’m a sucker for texture). I use the microcontrast one and very seldom other sharpness than zero. I’ve used it for years; recently made new tests with huge phase one’s files and despite that (for instance) progressive downscale+added sharpness produced more eye catching images (razor look), C3C was reassuringly consistent through very different textures, shapes, hues, etc.
Unfortunately you’re right, C3C only runs inside PS. It feels a bit like an insult in such a place as this, but - as pimps as adobe/apple/and the kind have become - I’m a PS user for almost 2 decades… and it comes a time when you just know it (the good and the bad) and so I get to my final image 10 times faster than in other platforms. With LR even faster. For my non paid work though I very much enjoy exploring and getting to know (and admire the work in) such powerful tools as GIMP, photoflow, GMIC, Darktable and so on.
I have my own aesthetic theories, borned from actual doing, about video being closer to painting and photography to drawing… but that’s for another day maybe 
Thanks also for the heads up, no-halo it’s what I normally use in GIMP , but again (and sorry for repeating) tested it against C3C and quite prefer the later… maybe it just has more to do with my heterodox workflow; I do appreciate that you pointed that out, peace.
You could also try using the PhotoFlow plug-in for downscaling the images, to see how it compares with other methods… the latest plugin uses the built-in resizing function of VIPS, which in turn is based on block shrinking + variable-sized lanczos3 (see here for some details).
Here are the settings I usually apply for web output:
@Carmelo_DrRaw PF downscale is super fast and very nice, there are some UI annoyances but I’ll address those (and some others, collecting, minning, collecting) in the PF thread; for nitpicking a tiny bit too sharp.
So, in order to have something to look at and for also for others to be able to draw their own retinoghts, I put together a super quick and dirty test, using all medium format files (hassy, pentax, phase one) from the good world wide web’s arquive. Below
Procedure: A 10075 x 5677px (300dpi) combo of images, converted to sRGB and downscaled to 15%.
In comparison PS’s own bicubic downscale is so bad that I threw it away.
GIMP No-halo was unbearably slow (maybe my parta’s version… dunno) that I did’t endure a low-halo export; not bad a result overall, 3rd best.
Photoflow VIP’s based algo is sharp, a little too much, it also introduced a gamma variation which I didn’t particularly like, darker dark tones (like if a ever so slightly local contrast) and lighter mid tones. If you look at the attached image, you’ll see a bit of moiré in 9 and 10. Second best by my book
C3C (contrast adjustment) with 0 sharpness managed the more pleasant result. The details are finer (higher resolution?) - see sign and greenhouse. If more more sharpness is desired 1 and 2 (the limit) at the expense of introducing artefacts and moiré. Still unsurpassed 
If anyone wants to DL the tests and the huge chart - https://drive.google.com/file/d/0Bxtrjp4jb-YsV18tVEZkS01oQm8/view?usp=sharing
CONTENTS
Hope admins don’t mind me gluing the site logo, if someone finds the image it would be clear that comes from these forums Case you don’t want it, I can upload and change it for a logoless one
Very interesting! I will point the VIPS developer to this comparison, maybe he will have some good idea for improvements…
Moreover, next PhotoFlow version will make it easier to do rescaling in linear gamma encoding.
Could you meanwhile maybe do a little test and rescale your reference image with photoflow in linear gamma?
You can add a “colorspace conversion” layer that converts to this linear Rec.2020 profile, do the rescaling, and then add another colorspace conversion layer to go back to sRGB.
I’m curious to see if the photoflow version improves by doing this…
Thanks!
You nailed it small details are better, more defined also moiré almost gone and gamma looks gooooood… that I can see just a side effect, a colour banding of sorts
Is this color banding also visible when downscaling in the gamma-encoded case?
How would you compare the linear gamma result with respect to C3C?
I’ve sent a message to the VIPS developer concerning the color banding…
Here is the link to the issue report, in case you want to follow the discussion: https://github.com/jcupitt/libvips/issues/536
By the way, how is the kitten doing?
the fucking bastard is chewing my hand!!! You’ve never seen animal hunger till you meet an starved blind kitten, definition of elastic: her belly. It’s overwhelming to see a tiny creature in such bad shape and still fiercely eager to explore, to play and to trust such a clown as I am. She loves this thread, nap time 
Thanks 4 the heads up on the VIPS’ issue
next PhotoFlow version will make it easier to do rescaling in linear gamma encoding.
looking forward lineardized pf-scale 
I did a quick test, and the banding seems to come from Jpeg compression and not rescaling itself… no banding when saving in TIFF format. Maybe you can try the same?
I did a quick test, and the banding seems to come from Jpeg compression and not rescaling itself… no banding when saving in TIFF format. Maybe you can try the same?
Yep, confirmed with 2 new tifs files (sRGB grayscale and colour gradient). Jpeg export introduced nasty colour banding, grayscale grad seemed almost unaffected. Tif’s export are pretty clean 
PD
Probably not relevant but I noticed the PF’s jpegs are considerably bigger than PS’ ones
Yes, I am aware of this and this is part of the stuff I am looking into at the moment. Jpeg output needs to be improved and better optimised in photoflow…
Taking the TIFF files as reference, I would you now judge the photoflow rescaling compared to C3C? Are we still significantly far off?
That reminds me that old joke,
I did run some more tests, and I’m pretty sure the overall gamma change is from c3c 
In comparison, and assuming that the gamma shift is indeed from c3c:
PF export of jpeg w downscale — has too big a (I’ll called it) radius sharpening (some moiré visible). In circular gradients, near the edges aliasing artefacts are present.
PF export of jpeg w downscale and the 2x colour conversion — better radius (still a bit of moiré) but introduces (colour) banding. In circular gradients, near the edges aliasing artefacts are present.
PF export of tif w downscale and the 2x colour conversion — much better radius (less moiré) no (colour) banding whatsoever.
PF export of tif w downscale — same as above.
C3C (either tif or jpeg) — hits the spot details wise but creates a slight lighter output.
So it seems that c3c it’s not perfect, introducing a gamma shift, also that VIPS’ lanczos 3 has (IMO) too big a sharpening radius and PF’s jpeg engine is not outputting as clean as jpeg as it could. I’ld say that despite not resolving fine detail as neatly, PF’s downscale (w tif) vs c3c is a draw 
PD
Today’s such a crazy day, it started with me cooking several meals, after breakfast I had already lost a tooth and short after an extra kitten, WTF!!! A beautifully fluffy - and I mean it, cotton candy hair - big headed male that looked like a mini grey wolf. Now I’m all alone in this big wear-house (buahhhh); we’ll see how t ends…
