Output changes using same saved .pto file

[J Harvey]

I am trying to manually create a HDR or mixed exposure image.  I want to stitch 8 image together with three different exposure and sharpening settings.

I output jpg set A.  create control points, define projection, and all other settings, optimize, and stitch producing output A in tiff and saving the parameter file .pto

I derive a second set of jpgs of the same spatial extent, but different levels and sharpening, and place them in the original folder, moving the A inputs to another folder.  The second set, B, are given the same name as the A set.

I open the original .pto file, and it loads the B set of images.  I make no changes, but go straight to 'stitch', and produce output B.  When I overlay B.tiff onto A.tiff, there are significant differences in the panorama.  The spatial extent is the same, but internally features do not align.  There are discrete regions with relative translations of image regions.  (ditto for subsequent versions as well, with variable differences)

What could be causing this, and is there a work around?  has anyone seen this before?

thanks

[Terry Duell]

On Sat, 16 Dec 2017 13:59:10 +1100, J Harvey <janethar...@gmail.com>
wrote:

> I am trying to manually create a HDR or mixed exposure image. I want to
> stitch 8 image together with three different exposure and sharpening
> settings.
>

[snip]

I may not have understood why you are using the approach you describe, and
wonder why you are not loading all the images and using stacks to get your
HDR.

Cheers,
--
Regards,
Terry Duell

[J Harvey]

I'm not trying to produce a traditional HDR image.  I have a panorama image with clearly defined foreground, mid ground, and distance elements, and I'd like each region to have it's own set of non-linear level adjustments, and consequently the sharpening and saturation can vary between them.  There is a higher level of variance that can be produced by producing three sets of jpgs from the RAW images, and making 3 panoramas from each set of images, rather than producing a middle level jpg series, making a panorama out of that, and then stretching the panorama independently for the foreground, mid ground, and distance.  

I have one .pto file for the panorama, and three sets of images that are named the same, in different folders.  But when I apply the same .pto panorama parameters file, with no changes, the output has somewhat significant spatial differences.  I could understand if the control points varied between each set of images, since image elements present slightly differently with different levels and sharpening, but I'm using the same set of control points.  

In fact, just running the same .pto file subsequently with different images gives me different results.  It's like there's a stochastic element to the panorama creation, it doesn't produce the same results, consistently, given the same inputs- I can't reproduce the same panoramas.

Linux, Hugin

Version: 2015.0.0.cdefc6e53a58

[J Harvey]

Lets see if this works. 

I opened a saved .pto, and ran stitch three times without changing anything, and it produced three different images.

 

 

[Marius Loots]

Hello All,

I have experienced the same problem. It relates to how nona, enfuse or enblend select the areas which is to be merged.
Can't remember at the moment exactly where the issue was. I revert to the 2012 version of hugin when I encounter this.
I will see if I can where exactly the change came. You can use masks to make each blend line the same for the same pto,
but it was too much of a pain and easier to revert.


--
Best regards,
Marius Loots

This message and attachments are subject to a disclaimer.
Please refer to http://upnet.up.ac.za/services/it/documentation/docs/004167.pdf for full details.

[Sean Greenslade]

On December 16, 2017 5:22:08 AM PST, J Harvey <janethar...@gmail.com> wrote:
>Lets see if this works.
>
>I opened a saved .pto, and ran stitch three times without changing
>anything, and it produced three different images.
>

><https://lh3.googleusercontent.com/-IwBg4xmFG0Q/WjUd93odqxI/AAAAAAAAB5c/-bL1OuVAe60kIhZovMiWuO30IU1KvTyegCLcBGAs/s1600/Hugin.gif>
>
>
>>
>>
>>

Enblend has the --save-masks and --load-masks option so that you can reuse the same source masks across different blends. But even so, I believe the actual blending seam (and seam width) is still driven by analysis of the input images, so I'm not sure if it's possible to get the same exact stitch with different image files.

--Sean

[David W. Jones]

Sounds like you're doing focus stacks.

This search returned a number of articles and such about doing that with Hugin:

https://duckduckgo.com/?q=focus+stack+hugin&t=ftas&ia=web

David W. Jones
gnome...@gmail.com
wandering the landscape of god
http://dancingtreefrog.com

Sent from my Android device with K-9 Mail.

[J Harvey]

Not a focus stack.  A 2 row panorama, same focal length, same exposure, roughly the same focus at infinity (landscape image).  

The panorama output is fine.  It's just not repeatable, even with the exact same input files, and no changes to the .pto file. I literally select 'stitch' twice (exposure fused from any arrangement, optimal size, no resizing set in preferences, all other values to default), three times, four times in a row without changing any of the variables, and each output panorama image is different.  See the gif I posted above.

I don't understand how the algorithm works, that the same exact variables produce different results, unless there is some random seed somewhere.

What I'd like to do, is to produce panoramas with consistently matched pixels, so that I can manually blend differently stretched jpgs derived from the same RAW files.

[David W. Jones]

Hmm, then maybe just output mapped images, since you don't want Hugin to
blend them anyway? Then you can blend them manually.

I don't understand the seam finding algorithms, either. For all I know,
NONE of them use a repeatable seam.

--