I know Hugin can take various partially overlapped photos and create
a single large image. The typical use is a bunch of side by side ones
for panoramas. I've had a bit of luck doing that (1 tall x N wide).

Supposedly it can also create images from M tall x N wide for M > 1,
N > 1, but I have not had good results.

I've got a photo stitcher on my phone (Bimostitch Pro) that can work
from video. It works okay for the case of "stand in one place and
turn left to right to scan horizon", but does have issues when too much
of the horizon is very similar. I run into that with coastal shots. The
on-shore stuff is fine, then the water bits start to go crazy.

But Bimostitch does not work with the case of "hold camera at one angle
and move", you know like a dolly shot. I've had people tell me Hugin can
do that, but I haven't figured out how.

Probably the ideal program could take a video with a zig-zag sweeping
view of a wall and reconstruct an image of the whole wall. I'm thinking
about the case where I can't get far enough back, or a wide-enough-angle
lens view of a mural to get the whole thing in a single shot.

Any other tools I should look at?

Elijah
------
Bimostitch Pro is neither free not open source, so no link

In article <eli$2106211943@qaz.wtf>,
Eli the Bearded <*@eli.users.panix.com> wrote:
>I know Hugin can take various partially overlapped photos and create
>a single large image. The typical use is a bunch of side by side ones
>for panoramas. I've had a bit of luck doing that (1 tall x N wide).
>
>Supposedly it can also create images from M tall x N wide for M > 1,
>N > 1, but I have not had good results.
>
>I've got a photo stitcher on my phone (Bimostitch Pro) that can work
>from video. It works okay for the case of "stand in one place and
>turn left to right to scan horizon", but does have issues when too much
>of the horizon is very similar. I run into that with coastal shots. The
>on-shore stuff is fine, then the water bits start to go crazy.
>
>But Bimostitch does not work with the case of "hold camera at one angle
>and move", you know like a dolly shot. I've had people tell me Hugin can
>do that, but I haven't figured out how.
>
>Probably the ideal program could take a video with a zig-zag sweeping
>view of a wall and reconstruct an image of the whole wall. I'm thinking
>about the case where I can't get far enough back, or a wide-enough-angle
>lens view of a mural to get the whole thing in a single shot.
>
>Any other tools I should look at?
>
>Elijah
>------
>Bimostitch Pro is neither free not open source, so no link

Handbrake.
--
Member - Liberal International This is doctor@@nl2k.ab.ca Ici doctor@@nl2k.ab.ca
Yahweh, Queen & country!Never Satan President Republic!Beware AntiChrist rising!
Look at Psalms 14 and 53 on Atheism https://www.empire.kred/ROOTNK?t=94a1f39b
Changing the dictionary does not move us forward. -unknown Beware https://mindspring.com

In comp.os.linux.misc, The Doctor <doctor@doctor.nl2k.ab.ca> wrote:
> Eli the Bearded <*@eli.users.panix.com> wrote:
>> Probably the ideal program could take a video with a zig-zag sweeping
>> view of a wall and reconstruct an image of the whole wall. I'm thinking
>> about the case where I can't get far enough back, or a wide-enough-angle
>> lens view of a mural to get the whole thing in a single shot.
>>
>> Any other tools I should look at?
> Handbrake.

I'm looking at the docs, and I don't see how to make a still stitched from
multiple frames. Where do I find out how? Maybe you have a sample
command line or other documentation to point to?

https://handbrake.fr/docs/en/1.3.0/

Elijah
------
knows handbrake only as a transcoder

On 6/22/21 2:34 AM, Eli the Bearded wrote:
> I know Hugin can take various partially overlapped photos and create
> a single large image. The typical use is a bunch of side by side ones
> for panoramas. I've had a bit of luck doing that (1 tall x N wide).
>
> Supposedly it can also create images from M tall x N wide for M > 1,
> N > 1, but I have not had good results.
>
> I've got a photo stitcher on my phone (Bimostitch Pro) that can work
> from video. It works okay for the case of "stand in one place and
> turn left to right to scan horizon", but does have issues when too much
> of the horizon is very similar. I run into that with coastal shots. The
> on-shore stuff is fine, then the water bits start to go crazy.
>
> But Bimostitch does not work with the case of "hold camera at one angle
> and move", you know like a dolly shot. I've had people tell me Hugin can
> do that, but I haven't figured out how.
>
> Probably the ideal program could take a video with a zig-zag sweeping
> view of a wall and reconstruct an image of the whole wall. I'm thinking
> about the case where I can't get far enough back, or a wide-enough-angle
> lens view of a mural to get the whole thing in a single shot.
>
> Any other tools I should look at?
>
> Elijah
> ------
> Bimostitch Pro is neither free not open source, so no link
>
Have a look at this: http://www.dojoe.net/tutorials/linear-pano/

Which was found on http://hugin.sourceforge.net/tutorials/index.shtml

In comp.os.linux.misc, dillinger <dillinger@invalid.not> wrote:
> On 6/22/21 2:34 AM, Eli the Bearded wrote:
>> Any other tools I should look at?
> Have a look at this: http://www.dojoe.net/tutorials/linear-pano/

That's from 2009, hopefully things are a lot better since then because
that's a very intense process.

Selected quotes (with no elipses):

After that, the fun part began. PanoTools can't move pictures in the
way of assigning an X and Y offset to each of them. Instead, each
picture can have an individual "center shift" in the "Camera and
Lens" tab. I made shameless use of that by simply unchecking the
"inherit" box for the d and e parameters.

Now, when optimizing, Hugin didn't allow me to exclude my anchor
from having it's lensshift optimized because it groups these
settings by lens number, so I had to assign a unique lens to my
anchor by selecting it and then clicking the "New Lens" button.
After that, I could move to the next tab.

Autopano-SIFT generated control points for putting two adjacent
images together, and it did a really nice job there. But! Optimizing
with only these control points gives Panorama Tools no hint
whatsoever about which parts of the images should be on the same
height, or to say it more technically: PTOptimizer has too many
degrees of freedom yet. With over fifty pictures to put together,
small errors will accumulate and you won't end up with a straight
picture of your wall, but with a pretty snaky stitch...

When the final stitch was finished and proudly lay on my harddisk in
a 600MB PSD, the actual fun began. And by fun, I mean four hours of
mind-boggling, eye-straining precision mouse wrangling1.

I opened the file in the GIMP and, one by one, started optimizing
the layer masks so that the transitions lay inside single-colored
areas most of the time, which meant following the shapes of the
graffiti, sometimes over pretty long distances. At this point,
having more than 50% of overlap saved my life several times.

Let me elaborate on the "enblend messes up the transition lines"
part: What enblend does is generate a transition line that's
farthest from the borders of the overlapping section. Thus, that
line is not dependent on actual image content, but rather on the
geometry of the remapped source images. Actual image content is only
considered during the actual blending, where enblend basically
adjusts the transition line's feather width based on image content.

> Which was found on http://hugin.sourceforge.net/tutorials/index.shtml

Elijah
------
still wonders if it is pronounced hu-gin or hug-in