Yes @stoffball
9 rows of 3 bracketed pictures… Stack 4 has been selected for AC
Funny thing is the first 3 stacks are OK, then 1 or 2 shifted, 1 or 2 OK, rest shifted again…
Sorry, but this is not visible in the screenshot. You are showing the general information and photometric parameters. I was speaking about geometric parameters (image positions). Can you attach the pto file (images are not needed, alternatively can you provide a download link)?
Please find the file below.
https://drive.google.com/file/d/1kFWy12M7fETT65HwZxMSsLBQ8f8FAlA_/view?usp=sharing
@Henk
ok, the image position are linked. But there are 2 other serious problems with the project file.
First, I doubt that you shoot the images with a lens with cylindrical projection. Switching back to a rectilinear lens projection decreases the control point error significantly.
Second, the overlap between your images is about 60-70 % and not 30 % as you written.
This interacts with the automatically stack detection and Hugin will try to merge 2 stacks into 1 HDR image. see Hugin Stitcher tab - PanoTools.org Wiki from the Hugin wiki, also included as help in Hugin. Even if is written under caption Exposure fusion the same algorithm applies also to HDR output.
Quote
If Exposure fused from stacks (added comment for HDR) is enabled then hugin will group the input images into exposure stacks by comparing positions, any images with more than 70 % overlap are grouped like this. Each of these bracketed exposure stacks will be exposure fused with enfuse (merged with hugin_hdrmerge into a HDR image) and the results seam blended together into a panorama with enblend.
To fine tune this read the last note in the list in the above mentioned file:
Quote
For special cases the percentage for decision for the overlap can be fine-tuned (only in expert). Go to Photos tab and select group by output stacks and change the minimum overlap value. If you want to use the already assigned stacks instead of the deduced stacks, set the minimum overlap value to -1.
PS: This behavior the area dates back before explicit stack handling in Hugin was implemented. Changing this would break backward compatibility.
1 Like
Thank you for your efforts @stoffball !
I assume that resetting this now would mean all current control points will be invalid. Might try that next weekend or week… The lens was/is a 50 mm and I already was surprised to see the deviations near the edges but attributed it to roll and not 100.00% level camera rotation.
What Hugin said at some time ago to me was:
“Since the vertical field of view is not too wide, you could try setting the panorama projection to cylindrical Cylindrical projection preserves vertical lines, unlike equirectangular.”
Now I assume that that only applies to the stitch tab 
As there are many high rise flats in the distance, vertical lines need to be preserved. As there is water along the lower edge, and a dark sky along the top edge, there is only narrow band in the middle (only larger directly in front of me). So overlapping around 50% was my target. I (mis)understood that 30% was the minimum…
(beep (beep)! Never knew this!!! I already wondered where the 5 (instead of 9) stacks in the log file came from. Output stacks =/= input stacks… Hm so how to change that.
Set to -1 and hit Re-Optimise… Works
Bit of patience…
The result is absolutely gorgeous!!!
Thank you again!!
1 Like