 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | You are not logged in.
Some time ago I wrote about tufuse, but now that enfuse got released as stable I am more excited over it because not only is it open source, but it runs much faster and allows me to process larger images.
I wrote a script for running enfuse on a batch of files, its partially based on the old tufuse.sh script, but I rewrote and changed a lot of it, it should give you more control now and not ask redundant questions, e.g. when you select 3 images its quite obvious that you want to have all 3 of them in a group, so asking for that would be redundant.
You can get the script and read the documentation and how to set it up here:
http://www.panopixel.org/bashenfuse.php
For my needs this is enough, but if you have any suggestions then please write here. For example I might add support for auto image alignment in the future, but then again this seems like a waste of time since we have APP for that, time that could be better spent photographing and eating :]
For those interested, enfuse allows for tweaking some 'advanced parameters'. I modified the script to test them all and I found that the defaults are in fact best for most situations. The only parameter you might want to sometimes experiment with is --HardMask, it gives sharper edges around moving things, but it wasn't worth including in the script as a yes/no options since most of the time it would be a waste of 1 click and just annoying. If you want to try it out, find this line:
enfuse -m 1000 --compression="$compression" $blend180 -l "$levels" -o "$output" "$@"
and add --HardMask before the -o
Of course the dream Enfuse GUI is a cross-platform program that allows tweaking all parameters on one screen and shows a miniature live preview but that's beyond my powers at the moment :]
Offline
Great, nice idea 
I like the spirit and the way it has been done.
Offline
Great! Thanks for sharing that.
Offline
I'm not following the whole cmd line tools development, so could someone enlighten me - is Enfuse doing the same job as Tufuse (exp blending, fake hdr out of raw files, focus blending)? Does Enfuse require the files to be pre-aligned (via APP or other software)? What happened to Enblend?
The pre-alignment requirement pretty much makes those tools useless to me. I cannot get a perfectly aligned layers out of handheld panoramas...
Offline
Maybe the script could also use align_image_stack?
http://wiki.panotools.org/Align_image_stack
Offline
Thanks
[bo] yes, enfuse now does the same as tufuse. Enfuse does require files to be aligned properly, but I could use something like what fma38 suggested, align_image_stack, to automatically align them before stitching... just thats kind of pointless since we have APP, so its APP's job to align the handheld different exposures into the same panorama and then render using layers and enfuse the layers afterwards... If you want to enfuse your handheld source images then align_image_stack would help... is it that important?
I did some tests this year and I decided that *fusing/tonemapping the whole panorama (if possible, since if its too big then its impossible) is better than doing that on the source images, here is the thread: http://www.autopano.net/forum/t4245-why … -stitching
*fusing/tonemapping the source images produced a more fake-hdr-look and it caused big problems around bright areas like the sun.
klausesser you are a supporter of *fusing/tonemapping (from now on I'll just call it fusing) before stitching, why? I would do that for 2 reasons only:
1- you can fuse source images no matter how big the final pano will be.
2- APP-1.x doesn't choose the same place for merging from one photo to the next in a pano which consists of several layers, each having a different exposure. This can result in small ghosting artifacts in a pano even if the source images were perfectly aligned. I hope APP-2.x will fix this with its notion of stacks.
If you have any other reasons I will gladly read them!
Offline
Thanks for the info.
My handheld panos cannot be processed with APP + cmdline Tufuse, I've tried that. The main reason is I don't do bracketing and shoot in a fashion that's not really usable for that workflow.
A classic example is a landscape with bright sky and darker forest below (around me). I make a simple 2x2 horizontal shots (2 sky and 2 land), so I can get a relatively square stitch. What should I do next?
Here's what I tried several times and gave up:
1. Make a pano in APP with two layers, land and sky. Each layer is color corrected, each with yellow anchor.
2. I render two layers, so I get two TIFF files. Those two files *do not* match *perfectly*.
3. I run tufuse on those two TIFFs. Two problems arise.
--- not being aligned, there are some kind of ghosting artefacts, overlapping objects (usually branches), etc
--- in two of three cases, the file is just too bleak and no amount of curve editing can fix things
Any ideas?
Offline
Try this:
0- Set up your pano however you like
1- Make 2 layers, lets call them ground and sky
2- Go to color anchors, enable "Optimize by layers" and place a yellow anchor on each layer
3- Optimize color
4- In the Layers panel, click on 
R to reset layers
5- Now don't touch color correction anymore, just render your pano to 1 layer using smartblend to avoid ghosts. Thanks to the R trick you can get smartblend to render a final image consisting of only 1 layer, that includes those two Ground and Sky layers using different color correction for each of them.
This doesn't work well when there are extreme differences in exposure between each layer since you will get low quality bleached out color where APP made the light images darker and the dark images lighter, but it can be used successfully.
Other times I have to just render both layers, 1 per file, and blend them myself.
Offline
fma38 wrote:
Maybe the script could also use align_image_stack?
http://wiki.panotools.org/Align_image_stack
I made a simple test using the align_image_stack script: it works really fine.
Last edited by fma38 (2008-11-30 10:11:22)
Offline
