 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | You are not logged in.
Pages: 1
This is a picture to test Autopano Pro. It's not a picture for beauty just for testing purpose (or fun, whatever).
So, I need big panorama. This one is a 314 pictures panorama done with a 300mm with a manfrotto head, resulting in a 626 Megapixels panorama.
It tooks around 25 minutes to shoot and less than 10 minutes to stitch.
For the render time, it depends on options.
- First test : multiband, bicubic :
Rendering : 1h
Cache folder used : none
url : http://www.gigastitch.com/challes_300mm_multiband.html
- Second test : smartblend, bicubic :
Rendering : 6h (in debug, I forgot to launch the release version)
Cache folder used : up to 12 Gigas ( ouch ... )
url : http://www.gigastitch.com/challes_300mm_smartblend.html
Offline
10 minutes stitching is incredible! (...though I believe you.) Printing could be the slowest part (5 or 6 m² - 60 sq. feet)
Offline
Yes, I was investigating the printing part. At 300 dpi, it's around 1.23m high x 3.81 meters long ...
Offline
Adding some smart-sharpening and local contrast would vastly improve the net result. Of coourse photoshop might need as much stime as the stitching process.
Offline
Photoshop didn't open the jpeg ... (over 30000 pixels wide). I have to convert the jpeg into tif for beeing able to open it in PS (There's some limitation in jpeg support in PS).
Offline
Nice work!
Can you elaborate on your setup and shooting procedure?
For sure, you didn't need a nodal point adapter. Did you use a 3-way head or a ballhead?
I made preliminary tests with my Nikkor 180/2.8 (mounted in portrait orientation) and found that the horizontal rotation has to be done in steps of 3-4° and the vertical rotation in steps of 5°. This is very hard to do, especially with a ballhead. At 300mm, things are becoming even more critical.
How did you manage to shoot so exact and fast (only 5 seconds per shot)?
Best,
hadron
Offline
The shooting procedure :
- I used a manfrotto tripod and panohead, the spherical one. It's robust enough to handle the Nikkor AF 300mm F4 I have. I used my old D100 as a camera.
- The most critic part was to not miss a row in the scanning process (I did that once on a 300 pictures panorama ... ). So I used this rule, when achieving one edge, modify the pitch of camera to ensure that bottom half of previous row is now the top half of this row. It creates a huge overlapping coefficient average, 3, but I'm sure I didn't miss anything.
(For the calculation of the overlapping coef : 3 = 314 pictures * 6Mega from D100 over 626 Megapixels final).
- I don't measure angle exactly when changing the yaw of the camera. I prefer to look the camera to ensure a good rotation.
Offline
Boustrophedon that is! When reaching one edge change pitch (up or down) and change direction (left to right then right to left, etc.) This is like plowing a field and the way very ancient greeks used to write.
Using a 300 mm (equivalent) I found this kind of pano is possible with a decent tripod and no pano-head, but I would hate having to use a ball head.
That the rotation axis (yaw) is vertical matters when horizontal pano FOV is large. If cylindrical projection is used one have to decide whether the horizon will be straight or curved (some streets or river banks could be bent when the horizon is straight).
When rectilinear projection is used, then more images are needed near the top and bottom at the middle of the view (or the top and bottom parts near the edges must be cropped.)
Offline
Pages: 1