[Play Raw] First foss processing - panorama

I had to crop the edges of each HDRmerge stack to obtain a halfway decent Hugin pano. Then a film simulation in RawTherapee.

CRW_3486-3488 - CRW_3492-3494.tif.pp3 (11.9 KB)

I blended some of pat’s film simulations in GIMP, but I don’t remember the exact recipe unfortunately.

:+1:

I used this processing in RT5.4 and exported the 3488, 3491, 3494 files.
CRW_3488.tif.out.pp3 (11.0 KB)

Then stitched them using the Image Composite Editor (You can also do it through GIMP) and loaded panorama in GIMP. In GIMP I processed it to taste. I remember increasing the saturation a bit, darkening the sky a bit. In the end I just put rectangular selection around the Church, and brightened that part using the Levels tool and then inverted this small selection and darkened the rest of the photo. I noticed that the panorama had some artifacts in the clouds so using the clone tool removed them. That’s about it.

Your version is very good. The details one can see (in your version) are mind boggling!
Thanks for the photo. It was nice to work on it. Can you tell us which city is this? In fact, it might be better if you put the name of the city in the title itself.

My take. I also only used a single exposure per shot. Preprocessed in darktable for demosaicing and color denoising, then stitchted in Hugin and modified afterwards in darktable.

1 Like

Rio de Janeiro, this is the exact spot from where I took the shots:

image

I think I won’t change the title, let it mark my first.

@paperdigits You got the most natural one, followed closely by @Thanatomanic, @agriggio and @age. :slightly_smiling_face: These are the colors that I usually see in there, your sky is not so contrasting as mine, and it’s much closer to what it was at time of shooting.
If you zoom in the trees on ths slope at left, or the balcony at right, you can see halos.
By using the EV -1, you probably got much noise, which required strong denoising and sharpening, right?

Yes, I applied a fair amount of denoise. I probably under sharpened just a touch to keep it looking natural.

Put it on a tripod with a pano head to get rid of the halos :wink:

I started using hdrmerge, as it seems the tool everyone uses for hdr, and, by increasing the Mask Blur Radius from the default of 3 pixels to 10, I could almost get rid of those artifacts:


Maybe not clearly seen in this screenshot, but when I switch windows between the default and the 10 pixel radius, I can clearly see that the artifacts are almost gone, and everything else in the picture seems to be preserved.
Thanks for pointing that.

@everyone who used Hugin: what were your settings? Since I don’t have much experience with it, I would like to learn from you all :slight_smile:. These are mine (pto.zip) ATM.

@afre stack0_0 - stack2_0.pto.zip (3.0 KB)
Note: In Feature Matching, instead of CPFind, read CPFind + Celeste. Not sure why it didn’t save that…

Indeed, some of the settings weren’t saved in the pto. FYI, I did choose CPFind + Celeste because of the helpful parenthetical note.

So far, I am only aligning the brightest frames and already getting some artifacts in the towers and trees. I am not ready to share just yet :sunny:.

Could you paste a screen shot of those artifacts?

EDIT: Now that you mentioned, it’s true, I got at least one, this tower (on the left, the original, on the right, Hugin output):

Here is my first take. I didn’t use HDRMerge because I had problems with it. As usual, I took the low-level approach to processing. The scene itself is so interesting that I decided not to add much creative flair. Hence, it may be a bit bland for your tastes. Enjoy!

1 PhotoFlow → linear Rec.2020 without clipping (+EV only).
2 gmic → interpolate bad pixels → brighten (copy).
3 enfuse → recover contrast in sky from originals.
4 gmic → discard alpha → interpolate bad pixels.
5 hugin (middle frame as anchor) → CPFind+Celeste → geometric (all) → photometric (ldr,vwb).
6 gmic → discard alpha → interpolate bad pixels → brighten.
7 pnmclahe → moderate contrast enhancement.
8 gmic → recover chroma → resize → sharpen.


PS As you can see, I solved the artifact problem that I asked about earlier. My problem was that I clicked on the Create control points more than once generating an unhealthy amount of control points!

What I am still not satisfied with is the horizontal stretching at the left and right extremities, and also the cropping compared to other submissions. I guess that would be due to the centre frame being the anchor, and also the fact that I am only using 3 frames.

Impressive! Very natural indeed and yes, a bit washed out especially on the houses and buildings. The sky is exactly what it was at the time of my shooting, but I would add a bit of contrast, even if it’d make it kind of artificial. (I think I always see clouds and sky with more contrast than there really is.)
UPDATE: It’s also impressive how much noise you removed!

What do you mean by that?

What is that? I didn’t find it in PF?

This is in Gimp, right? But where do interpolate bad pixels is? Where do brighten (copy) is

Did you use a mask?

Where do discard alpha is?

Is this executed from terminal?

Where is chroma recovery?

Sorry flooding you with so many questions, but I’m totally noob in gmic, and I would like to reproduce your steps, in my quest to consolidate a floss workflow.

@Carmelo_DrRaw needing to correct the bad pixels after PhotoFlow processing would mean, PhotoFlow didn’t correct them when reading the raw file. That was also the case for RT until this fix. Maybe worth to implement something similiar into PhotoFlow :wink:

2 Likes

Not sure now, but when I open the original raw file in PF it doesn’t show any bad pixels (and there are a lot spread all over the image!):


In RT I agree that I needed to turn on Impulse NR to get rid of them

Ok, then it’s not an issue in PF. I just tought it was, becasue of the processing @afre mentioned.

With the fix I mentioned you don’t need impulse NR.

1 Like

The washing out is a side effect of local contrast enhancement or tone mapping. See @agriggio’s. I spent too much time on the panorama so I let that go.

As for global contrast, yes – it would benefit the image creatively.

The amazing thing is that I didn’t do any explicit smoothing whatsoever! I just kept a close eye on noise as I processed the image. Bad pixel interpolation and resizing, and using the brightest photos, may have helped but everyone does the first two.

Avoiding one-click solutions where I can (nothing wrong with those) and using the power of math, stats and friendship :rainbow:.

I didn’t use all 9 photos, only the 3 brightest and disabled clipping.

I don’t remove hot pixels or clip negative values in PF. I deal with them in G’MIC where I can use math, stats and extreme pixel peeping to determine which strategies are best. @heckflosse

I took the linear TIFs from PF, brightened and saved copies of them, creating artificial bracketed sets of 2. After brightening, the sky is perceptively featureless.

No. I merely maximized the weight to favour contrast. In retrospect, the previous step did that for me

Alpha is just another channel, which you could delete using any app including GIMP. enfuse and hugin use and export it. The reason I remove alpha is that I don’t need transparency or masking.

Yes, it is a terminal app. I talked about it in another thread. Just search the forums.

Chroma is the C component of LCH, which is the polar projection of LAB. You could modify it in GIMP, RT, dt, etc.

No problem. Clearly, my process is atypical. No other raw player does it this way, probably because it is a time waster. I would advise against it.

1 Like

I’m not sure if I will follow your advice :slightly_smiling_face:. After all, yours is one of the best results. One thing is for sure: I must overcome the fear of using gmic and its overwhelming number of tools.
I don’t want to abuse from your good will, but it would be great if you do a tutorial, with screen shots and the like, on how you’ve processed these images. You could name it something like A Conceptual Approach to Image Development,

Version 2 of my first take, improving my approach to step 7 and adding a slight s-curve to step 8. Enjoy!

2 Likes