<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html;charset=ISO-8859-1">
<title></title>
</head>
<body text="#000000" bgcolor="#ffffff">
For what it is worth, I get good results with Autopano 1.02 Alexandre
Jenny with following parameters:<br>
<br>
autopano /keys:20 /search:2 /ransac:1 /size:1024 /ptgui /nomove /noclean<br>
<br>
This is in a workflow that is typically<br>
Canon Digital Rebel at large/fine JPEG resolution<br>
Pano12.dll : 2.7.0.3 Jim Watters<br>
Autopano 1.02 Alexandre Jenny<br>
PTGui with PTStitcher<br>
Enblend 1.3 Andrew Mihal<br>
<br>
I recall that it took some messing around to get good parameters for
Autopano 1.02. <br>
<br>
Like others I never succeeded in getting autopano-sift to work under
Windows.<br>
<br>
--Rik<br>
<br>
Mike Runge wrote:<br>
<blockquote type="cite"
cite="mid200406240700.i5O704i05642@ks408.kasserver.com">
<pre wrap="">Hi Ian,
thanks for sharing you workflow in that detail.
My setup:
Elitegroup K7S5A
AMD Athlon 1400
1024 MB RAM
OS: W2K, SP3
Panorama software:
Pano12.dll: v2.7 Beta3 - Rik Littlefield
Autopano 1.02 Alexandre Jenny, like shipped with hugin package
Hugin: hugin snapshot - hugin_2004_03_25-08_41_win32
Enblend: Enblend1.3- with patch by Edouard Gomez as incorporated into the
latest enblend at <a class="moz-txt-link-freetext" href="http://wurm.wh-wurm.uni-ulm.de/~redman/enblend.exe">http://wurm.wh-wurm.uni-ulm.de/~redman/enblend.exe</a>
I didn't managed to compile autopano-sift on my own. I treid several times
the autopano.exe shipped with hugin, but only with bad results. Even autopano
did not incule all images belonging to the pano or it finds to much unwanted
points. I also have difficulties with the created pto-file, where each images
seems to have its individual lens. I corrected that editing the pto-file by
hand, but till now I never managed to get good results using autopano.
Would you be so kind to share the autopano-sift executable with me so that I
can try this one?
Agreed on all steps, I normally do 11 before 10.
I totally agree on seperating optimisation steps.
How do you do 14a and 15?
I understand, that in the moment it is nessesairy to remove the blank 1 Pixel
column, but how do you save that in gimp preserving the multiple layers for
the later enblend run?
16:
If I understand to right, you can edit the multilayer Tiff and save it back.
Could you step back to 14a and do some tweaks (e.g removing ghosts) before
running enblend? I understand, that some of the nessesairy tweak will be
identifyable after the enblend run and that you want to avoid running enblend
again. And I still don't understand how you remove the 1 pixel column and
save back the file?!
best, mike
=========================================================================
</pre>
<blockquote type="cite">
<pre wrap="">1. Copy the images I want to use for the panorama into my image working
</pre>
</blockquote>
<pre wrap=""><!---->folder
</pre>
<blockquote type="cite">
<pre wrap="">(C:Program Filesautopano-sifttmp)
2. Run a batch file I have created
(C:Program Filesautopano-siftbinAutopanoIt.cmd) to create the Hugin Project
File (normally Hugin.pto). The batch file checks to make sure that I have
</pre>
</blockquote>
<pre wrap=""><!---->got
</pre>
<blockquote type="cite">
<pre wrap="">some files to work with, creates the autopano key files and then calls
autopano-sift to create the hugin file.
3. Start Hugin and then open the Hugin.pto file.
4. Change the image hfov to my camera's setting (normally 38mm), set the
reference image for position (and exposure), and create vertical line
</pre>
</blockquote>
<pre wrap=""><!---->control
</pre>
<blockquote type="cite">
<pre wrap="">points for images (if I can), and add any additional control point pairs
</pre>
</blockquote>
<pre wrap=""><!---->where I
</pre>
<blockquote type="cite">
<pre wrap="">think they may be of some use.
5. Optimise for the default setting of "Positions (pairwise optim.,
</pre>
</blockquote>
<pre wrap=""><!---->starting
>from anchor)", then save.
</pre>
<blockquote type="cite">
<pre wrap="">6. Look at the "points" table and see if there are any wild points with
</pre>
</blockquote>
<pre wrap=""><!---->large
</pre>
<blockquote type="cite">
<pre wrap="">distances. If there are then review the point pair to see if is correct and
decide to adjust it, remove it, or keep it. If I changed anything then re-do
</pre>
</blockquote>
<pre wrap=""><!---->the
</pre>
<blockquote type="cite">
<pre wrap="">optimisation step.
7. Preview the image in the preview window, check that it looks OK, then
</pre>
</blockquote>
<pre wrap=""><!---->centre
</pre>
<blockquote type="cite">
<pre wrap="">the image and save the project.
8. Optimise using custom optimisation - for Yaw roll and pitch (yaw fixed
</pre>
</blockquote>
<pre wrap=""><!---->for
</pre>
<blockquote type="cite">
<pre wrap="">the reference image).
9. Optimise again adding barrel.
10. Optimise again adding view v.
11. Optimise again adding distortion a and b.
(Sometimes I get better results by swapping steps 10 and 11)
(I separate these optimisation steps because it seems to be a more reliable
</pre>
</blockquote>
<pre wrap=""><!---->way
</pre>
<blockquote type="cite">
<pre wrap="">to get a "good" solution without drifting off to a wild set of parameters.
</pre>
</blockquote>
<pre wrap=""><!---->After
</pre>
<blockquote type="cite">
<pre wrap="">each step I look at the preview window and the points table to see if
</pre>
</blockquote>
<pre wrap=""><!---->anything
</pre>
<blockquote type="cite">
<pre wrap="">has gone wild, and if it has then I revert to the previous step and try to
</pre>
</blockquote>
<pre wrap=""><!---->find
</pre>
<blockquote type="cite">
<pre wrap="">what went wrong)
12. Open the Hugin Stitcher Tab and set the pixel width to 1024, image
</pre>
</blockquote>
<pre wrap=""><!---->format to
</pre>
<blockquote type="cite">
<pre wrap="">JPG (90%) and then stitch now!
13. I then open this image (using irfanview) and check that it looks OK. If
</pre>
</blockquote>
<pre wrap=""><!---->it
</pre>
<blockquote type="cite">
<pre wrap="">does not then I try lots of things - more control points, less control
</pre>
</blockquote>
<pre wrap=""><!---->points...
</pre>
<blockquote type="cite">
<pre wrap="">14. Assuming it looks OK then I set the pixel width to the default (click
</pre>
</blockquote>
<pre wrap=""><!---->on
</pre>
<blockquote type="cite">
<pre wrap="">"Calculate Pixel Dimensions") and stitch again using nona and multiple Tiff
</pre>
</blockquote>
<pre wrap=""><!---->with
</pre>
<blockquote type="cite">
<pre wrap="">the file name nona.tiff saved into the folder C:Program
</pre>
</blockquote>
<pre wrap=""><!---->Filesautopano-sifttmp
</pre>
<blockquote type="cite">
<pre wrap="">14a. If the image is a 360 deg panorama then (since I started using the
</pre>
</blockquote>
<pre wrap=""><!---->patched
</pre>
<blockquote type="cite">
<pre wrap="">enblend) I open the created tiff file in GIMP and resize the image by one
</pre>
</blockquote>
<pre wrap=""><!---->pixel
</pre>
<blockquote type="cite">
<pre wrap="">column to remove the blank column of pixels at the right side of the image
15. I delete the 0 byte file nona.tiff (leaving nona.tif for later use)
</pre>
</blockquote>
<pre wrap=""><!---->then
</pre>
<blockquote type="cite">
<pre wrap="">start enblend using a batch file I have created
C:Program Filesautopano-siftbinEnblendIt.cmd.
The batch file looks for tiff files in the directory C:Program
Filesautopano-sifttmp and runs enblend to create a JPG file enblended.tif.
This works not only with multiple layer tiff files, but also with multiple
</pre>
</blockquote>
<pre wrap=""><!---->tiff
</pre>
<blockquote type="cite">
<pre wrap="">files (I have not tried it with a mix of both). It then uses image magic
"convert.exe" to convert this to a (smaller) PNG file and deletes the
</pre>
</blockquote>
<pre wrap=""><!---->enblended
</pre>
<blockquote type="cite">
<pre wrap="">tiff file.
16. I preview this then do any final "tweaks" using Gimp. If there are any
</pre>
</blockquote>
<pre wrap=""><!---->very
</pre>
<blockquote type="cite">
<pre wrap="">bad join problems (I mostly take hand-held shots and sometimes it's not so
</pre>
</blockquote>
<pre wrap=""><!---->good)
</pre>
<blockquote type="cite">
<pre wrap="">I use pieces of the images from the the Hugin output tiff file to paste
</pre>
</blockquote>
<pre wrap=""><!---->over
</pre>
<blockquote type="cite">
<pre wrap="">with corrections.
regards,
Ian Sydenham
</pre>
</blockquote>
<pre wrap=""><!---->
</pre>
</blockquote>
</body>
</html>