Hi, I would like to share this attempt to do an hdr from three exposures.
I used three images from the five ones attached in this post: -2, 0, +2 (3008, 3010 and 3012 files below)
My workflow:
1 - Darktable - I used the default setting and saved tifs
2 - align_image_stack - I aligned tifs from previous step
3 - Gimp - I opened aligned tifs from previous step as layers. I used the 0 EV image as background layer. Then I created luminosity masks from the 0 EV image. I used the MM channel to mask the +2 EV and the DD (or DDD, not sure…) to mask the -2 EV.
4 - Darktable - Final touches.
I basically followed the Basic Landscape Exposure Blending with GIMP and G’MIC tutorial by Ian Cylkowski, as well as Pat David’s article on luminosity masks.
The main issue with the final result is the appearance of halos in the top tree (which, I believe, are related to the fact that this was another hand held shooting and there are some alignment issues). Although I followed Ian’s steps to prevent halos, I couldn’t get rid of them in this image (clearly seen at 100% zoom).
I tried to attach gimp xcf file, but it is 500 MB.
Finally, I’m not sure how you will see this image, because I’m working on a low end Dell laptop, no monitor calibration at all. Opening the images on a Lenovo ThinkCentre desktop show them with a strong green hue, especially on which should be a yellow wall at left.
I hope you guys find these images interesting to work on.
Btw, this is Rio de Janeiro, Brazil, near downtown.
Have you considered using a tripod or something to stabilize your camera? There is no reason why you couldn’t as you aren’t taking action shots or have a time limit. I am a beginner in terms of taking actual photos (that is why I spend most of my time learning about post-processing). What I can see from my limited experience is chromatic aberration and lens distortion being the cause of the artifacts. Also maybe having a polarizing filter could help remove the glints on the trees. I know you are using a low end camera but there is a lot you could do to improve your shot .
I totally agree with @paperdigits, there’s plenty of light spill, and like him I’ve also taken the darkest 3012 into photoflow. I’d probably denoise it a bit in gimp and add some MC, sharpening and a bit of grain, but unless you’re aiming for huge size reprodction you’ve got legs to walk. BTW I like the framing but is missing a crossing the street. Tudo joia!
Very nice shot. All the verticals are indeed vertical, interesting composition and exposure is spot on! Pity that this sensor has so many dead pixels (last time too, there were many). Surely, a better camera than the one used is deserved by a person with such a photographic eye!
Anyway, here is an attempt purely with RawTherapee 5.4. Special attention was given so that the glare from the cobbled street is minimal. Thanks for sharing. CRW_3012-1.jpg.out.pp3 (12.4 KB)
Open DNG images to Darktable
I deleted all xmp data
Combine 3008, 3010, 3012 images with enfuse_pro_2.lua (Darktable filter that aligns images with align_image_stack and merges them with enfuse).
The resulting TIF image was opened with Darktable
Adjusted TIF image in Darktable, exported to JPG image
Yeah, most of the time I leave the tripod at home, as I’m not good at planing.
As for chromatic aberration, I agree that it shows up in the final image, but you can see there’s no issue with the images the others worked on. But I’ll check if I can reduce or eliminate it.
Yeah, I agree, but since I have this low end camera, I thought I could reduce some noise by using the overexposed images on the shadows.
@chroma_ghost What’s MC? Do you do all of this in gmic?
… I didn’t get it but… tudo joia
Humm… out of budget for now. Anyway, despite the dead pixels, I’m really attached to this one after I met CHDK and the way it “unleashes the power” of this photon collector unit. It really does!
Interesting, you don’t have the halos, which leads me to think that it was the luminosity mask thing I did which introduced them. I’ll try to tweak more.
Micro contrast, in the sense of achieving a bit more separation. Mostly g’mic, yes
4 movement, yellow buildings, green reptile. brazilian flag … also straight lines and human made vs organic shake, but maybe a tortoise’s better. i cannot but leave also this link
I do that too but the thing is that there are more ways to stabilize the camera than using a tripod. Be creative! I haven’t examined your raws yet but usually I find the first frame and the second one to have the largest deviation because you can move the camera when you press the shutter button.
One thing I found from your shots from the last PlayRaw that I just remembered is that the exposure of your next bracket set is taken from the previous set. Before you take the next bracket make sure that you readjust your exposure. That way you don’t accidentally make each set progressively darker. But it looks like this PlayRaw’s is just one set of 5.
@yteaot Yours is my favourite so far. It is less fake-HDR than your recent entries, more natural, which I like. The ghosting is mostly gone as well, esp. where the trees are, though on the left side we see some double imaged cables, etc.
Yeah, I know, but I remember clearly that I was in a hurry that day, the sun was cooking me and I was a bit concerned about safety, because that spot is not the safest one. Besides, I was reading at the time about the advantages of stacking not perfectly aligned images to get rid of noise (I know, the misalignment shouldn’t be that big, only a couple of pixels).so I really relaxed about stabilizing.
No, I don’t think so. Actually the bracketing is accomplished by a CHDK lua script, where you set up how many bracket exposures you want,the EV amount between brackets, and number of steps. It was coded by me (oneaty) based on another one, aiming bracketing and stacking. Each time the loop repeats, the camera starts from the 0EV exposure.
From this specific session I have a couple more sets, but they’re all badly misaligned so I left them out of this post.
Hi! This thread looked as a good opportunity to prepare another sample of scene-referred editing so I played with the files too
I used dcraw to produce linear tiffs as discussed in this thread then produced a simple stacking of three exposures (EV0 +2EV -2EV) using Blender’s nodal compositor.
The result is a scene referred image with wide dynamic range, so you have a lot of room to apply any creative grading afterwards.
…And now the link to the tiff files too.
The .blend file included in the compressed file with the images has a little bug I just fixed. Use the .blend file attached above.
I kept playing with the file and added a few more nodes. Now it has some unsharp-mask and clarity/local-contrast. Those operations were re-created using the existing compositing operations. For those interested, selecting the node and pressing tab enters the group and exposes the operations inside.
Hi @patdavid!
The blend file I just uploaded points to the three TIFF files created from the DNGs with dcraw.
They are rather huge. Should I upload them too? (I’m uploading them to my drive anyway)
Oh, and btw, you can use the same blend file for stacking your own shots too. It’s configured for stacking three shots only 0EV +2 -2 exposures but it can be modified for different values easily.
It only requires that the three input files are linear rec.709 tiffs produced with dcraw -T -4 from your RAWs.
@gez I don’t think that is necessary. As long as we know how to reproduce them. Maybe add the dcraw command in the actual post. Thanks for sharing something different.
@afre it’s all in the other thread I linked above.
Basically it’s using dcraw with the parameters -T -4 on your raw files.
I also use -w (for white balance), -n 100 (for some noise reduction) and -q 3 gave me good results, not sure whether it depends on the source files or not.
Trust the man page of dcraw better than me.
.
crossing the street. Tudo joia!
but… tudo joia
Forgotten… tropicalia alley dreaming of progress 
