My attempt with even more overkill
My eyeballs need a rest
Your words, my words. 500 years under such commandments (opulence, robbery, slavery), blessed by God, is a hell of an heritage, and people think it’s past, but I’m sure it isn’t. But let’s not digress.
Thanks for the fixes. I mixed English with portuguese, which is Barroco.
Regarding noise, the game was not fair for you, since I used four rounds of seven exposures 
I ran a 16 bit tiff output from Rawtherapee through the Nik DFine 2 plugin because it is automatic and I was lazy. I’m pretty sure you can tweak the noise reduction in Rawtherapee to get the same result, using detail recovery and curves. I like to spend my time on the colors and such.
1 Like
Went for simplicity. As usual, I use multiple apps: dcraw, gmic, enfuse and darktable. Power of friendship. 
1 dcraw → linear 16-bit TIF of darkest and brightest exposures.
2 gmic → interpolate unwanted pixels.
3 enfuse → weigh for contrast.
4 gmic → smooth (guided) → brighten.
5 darktable → semi-auto perspective correction → crop.
6 gmic → global contrast → slightly desaturate → local contrast → sharpen (tone) → resize → sharpen (edge).
Edit: The image above needs more contrast. Changes in italics. Here is v2:
1 Like
I edited my earlier post. I realized that the contrast was lower than I would have liked it. I also reduced the saturation. Hope you like version 2.
Could you elaborate on this?
I’m not suprised by the ocean blue balcony fencing, a color/pattern (the blue stripes) reminiscent of garments of the tribe of David, and its position aloft significant of specifically Mary’s genealogy.
1 Like
Nothing fancy really. The obvious light source seems to be from the skylights. So just thought that I will create a beam of light flooding the main statue and reaching the white cloth on the altar. The rest should be darker.
To implement this, I just duplicated the layer twice, made the top layer blend mode to screen. Put in a black layer mask and used gradient tool (FG TO BG) to apply it only on the left 2/3rd of the image. Applied that mask to the image and again repeated the same with right 2/3rd selected. When I apply it, the screen blend is visible only to the middle 1/3rd. Merged down this layer.
Repeated the same till I satisfied with the result.
Since foreground and the sides are darker than the top center of the image, I feel it gives a 3D effect.
1 Like
Many interesting and different interpretations, even showing that noise isn’t a bad thing necessarily 
Here’s my workflow:
1 - 4 x 7 x DNG → Hdrmerge - dynamic range compression of the seven different exposures → 4 x DNG
2 - 4 x DNG → Rawtherapee - Noise reduction and bad pixels → 4 x TIF 32 bit floating point and REC2020
3 - 4 x TIF 32 bits floating point and REC2020 → Gimp / G’mic → stack the 4 images (median) to reduce noise → TIF 32 bits floating point and REC2020
4 - TIF 32 bits floating point and REC2020 → Photoflow - generation of 5 images at 1 EV exposure steps: -1, 0, +1, +2 and +3 → 5 16 bit TIF REC2020
5 - 5 x TIF of 16 bits REC2020 → Gimp → Create masks to highlight / conceal light/noise, embedding them as alpha channels - 5 x 16 bits TIF REC2020 with alpha channels
6 - 5 x 16 bit TIF REC2020 with alpha channels → Enfuse → New dynamic range compression with more noise elimination → 16 bit TIF REC2020
7 - 16 bit TIF REC2020 → Darktable - various final adjustments (toning, curves, noise reduction, wb, exposure, local contrast, sharpening) → 8-bit JPG sRGB
My attempt using mostly Darktable.
-Applied a style I have for HDR preparation, basically removing the default sharpening and adding Base Curve, Highlight Reconstruction and Lens Correction.
-Denoise (bilateral) with radius 18, red and blue channels at max and green at default. This got rid of most of the noise.
-Copied this across all images.
-Exported all to Enfuse with the ‘pseudo hdr or dff image’ plugin as 16 bit tif.
-On the resulting tif: adjust white balance, Tone Curve for added contrast and Color Contrast for added saturation. Color Zones to boost the reds and blues a bit and Color Correction to reduce a blue cast in the blacks. Crop and Rotate to make the image level.
-Then I had some extra noise reduction in Neat Image (might be considered cheating since it is not open source software?).
-final sharpening in Darktable.
3 Likes
2 Likes
You might alleviate that burden by posting the input file to Neat Image 
Now talking serious, I’m really curious at how far you could reduce noise before Neat
@Thomas_Do I’m trying to load your sidecar file but nothing happens.
Is it a DT version incompatibility?
Mine is:
darktable --version
this is darktable 2.4.3
copyright (c) 2009-2018 johannes hanika
darktable-dev@lists.darktable.org
compile options:
bit depth is 64 bit
normal build
SSE2 optimized codepath enabled
OpenMP support enabled
OpenCL support enabled
Lua support enabled, API version 5.0.0
Colord support enabled
gPhoto2 support enabled
GraphicsMagick support enabled
OpenEXR support enabled
I definitely made the colours and noise excessive enough to hurt your eyes 
and to express my feelings on the subject of oppression and obtuseness of the architecture. Also, my take is a little reminiscent of tourists using flash 
when they aren’t supposed to. 
I’d say that is quite the adventure that you had there!
1 Like
I posted the wrong file. I already deleted the oringinal when cleaning up yesterday 
. Sorry!
1 Like
Very interesting image @gadolf!
I took the opportunity to test a new dynamic range compression module that I am adding into PhotoFlow (similar to the tone mapping module in Darktable). Here is what I have been able to obtain so far:
The steps are relatively simple:
- opened all images in HDRMerge and saved the result as a 32f DNG
- loaded the HDR DNG in PhotoFlow
- applied chromatic aberrations correction and blend HL reconstruction
- increased exposure by 2EV
- applied the new dynamic range compression module, followed by a filmic tone mapping curve (to compress the highlights) and an additional S-shaped curve
- rotation to straighten the framing
The processing is a bit “over the top” for my personal taste, but shows the potential of the algorithm. I will add the corresponding PFI file as soon as the new code will be committed to GitHub.
What do you think?
3 Likes