 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
Pages: 1
Hello!
I'm hoping that I could use panoramas to create wide angle shots with very narrow depth of field. I just ordered Nodal Ninja 5L for those and in addition to more tradiotional landscape/citycscapes. I'm pretty unexperienced with panoramas, I have few years photographed with 28mm and 45mm shift lenses to create panoramas and in addition done less than 10 "normal" panoramas/yearly.
Also would enjoy doing panoramas with 180mm macro and 300/2.8 tele - macro, close-up and medium distance with shallow(ish) depth of field plus landscapes. I'm not interested 360x180 stuff at all, just to create high resolution 2:1 - 5:1 ratio "fine art" photos.
Example of narrow depth of field panorama (Canon 85mm f/1.2 @ f/1.2 - 6 rows of 10 vertical images - Markins Q3 ballhead - had to left out the topmost row since I could not get the panorama to work if I included it - after a hard work of manual positioning of images and adding manually hard links to images I could make this one):
Link to 1920x746 size in which the narrow DOF is more obvious.
I would enjoy to create photos with this technique, but it seems so difficult to make the Autopano to understand how to stitch images which doesn't have anything sharp in some of the images. Positive surprise was that there were no parallax problems on foreground (like said above I just ordered panorama head - this was taken with normal ballhead).
I already tried, the "standard advice" gurus seem to give in this forum:
- manually add CPs by drawing rectangles (won't add any)
- adjust the CP detector quality (still won't add any)
In addition to 1.4.2 I tried with 1.9 beta version (mac version) - the beta did seem to handle this scenery better
Source images for this photo: here (zip - 52mb). I will remove them at some point of time, but there is no problem keeping them there for as long as this topic is having activity. Feel free to use images as you wish, they definetly don't have any artistic value...
--
Samuli Vahonen
http://www.vahonen.com
Offline
Hi Samuli,
If there is not enough information for sift or other stitching programs you will need a template. Kolor has a few ideas/ tools for the next release of AP giga. You should check out the gigapan import in the 1.9.1 beta of autopano giga. Shoot your images mimicking the default order of the gigapan plug in and give it a try.
Offline
gerardm wrote:
Hi Samuli,
If there is not enough information for sift or other stitching programs you will need a template. Kolor has a few ideas/ tools for the next release of AP giga. You should check out the gigapan import in the 1.9.1 beta of autopano giga. Shoot your images mimicking the default order of the gigapan plug in and give it a try.
Hi and thanks for giving hint about this gigapan import! I had shoot my test image by moving horizontally all 10 images and then adjusted ballhead vertical tilt and it seems Gigapan import assumes it done on other way around. Have to wait until I get Nodal Ninja, since with just ballhead it's too difficult to do "grid" vertically tilt first and then rotation. Also noticed "interesting" bug on 1.9.0 Alpha 1 Rev 3 (could not locate 1.9.1???) when you open File -> Import dialog there is no other way to exit the dialog except to create panorama, I had to force quit (equivalent of Windows world "go to Task Manager and end process") from Activity Monitor.
However if this just positions images to roughly correct places and then it uses SIFT the end result is not better. I can place the images manually to grid (takes just helv a lot of time) and then let the automatic things run. I may need to try some other panorama software, since for reason or another Autopano lacks functionality to manually place control points to pictures, which would be no problem to do, even it's blurred due to shallow depth of field there are in all images details which can be identified same as on neighbour images.
Autopano seems to have Save As and import for Panotools. In theory this would give possibility to do all what is automatically possible in Autopano and then Save As Panotools format, then define the other control points in other software and save, then import the modified file to Autopano and finish the panorama. Many years ago I used to have some Windows program (or Photoshop plugin, can't remember) which was using Panotools library, now it seems this product is no longer available/developed - any proposals for free (I don't want to buy two softwares for panoramas...) software which could read panotools format file, add/edit control points manually by user and save back to panotools format?
--
Samuli Vahonen
http://www.vahonen.com
Offline
Here is an attempt...
As you can see, I did not linked the 50 source images I stitched but only the ones where details were available. Unlinked "orphans" were set manually using the Move tools. This needs some practice but it was so easier than with my previous attempts to use this tools that I wonder whether this is because it was strongly improved or or because I did that for fisheye images (perhaps both.)
Anyway, as a new game is born: fuzzysaw puzzle 
Samuli, if you want the .project file, just ask for it.
Offline
A possible (but not certain) method could be to stitch together 4 wide angle shots (or more) and the long focal lenth shots, like I did there http://www.autopano.net/forum/t3526-pro … a-gapixels
Offline
Samuli
Maybe you could 'fake' it and get the effect you want.
See this:
http://www.panoguide.com/forums/galleries/5397/
Offline
GURL wrote:
A possible (but not certain) method could be to stitch together 4 wide angle shots (or more) and the long focal lenth shots, like I did there http://www.autopano.net/forum/t3526-pro … a-gapixels
Hmmm, this might help, or not since from blur the algorithm won't find anything. However this made me realize alternative tactics - take each photo twice, with f/1.2 and f/22. Of course this would only be possible with indoor scenes (like the example scene) since changing of exposure settings would take a long time between frames. Then these photo could be arranged to layers in Autopano and f/22 layer would be used for stitching. Interesting, however would not work most of the outdoor scenes with changing light.
GURL wrote:
Here is an attempt...
As you can see, I did not linked the 50 source images I stitched but only the ones where details were available. Unlinked "orphans" were set manually using the Move tools. This needs some practice but it was so easier than with my previous attempts to use this tools that I wonder whether this is because it was strongly improved or or because I did that for fisheye images (perhaps both.)
Anyway, as a new game is born: fuzzysaw puzzle
Samuli, if you want the .project file, just ask for it.
This is more or less the method I used myself, however for future use I would hope that there would be easier and less laborous than this method. Also even I spent many hours I could not get it exactly match. Also fuzzysaw puzzle was pretty unpleasant to make (at least to me - eyes start to hurt since trying to find details from blur...). You have less annoying bugs in manual placing of images than my version but there still are few. In this technique it won't hurt that panorama is taken very accurately, with even spacing between images.
mediavets wrote:
Samuli
Maybe you could 'fake' it and get the effect you want.
See this:
http://www.panoguide.com/forums/galleries/5397/
I don't want tilt effect - I want isolation of subject from background, example: Carl Zeiss Macro-Planar T* 2/100 ZF @ f/2:
If I would have taken this with 300/2.8 @ f/2.8 instead and created panorama the subject would stand out much more than it does now, also making it possible to print larger, I often print to 75cm x 50cm to display my photos and it's rather demanding to the image and Canon 1DmkIII with it's 10Mpix can barely do it, would not mind 50+Mpix... Another example with Carl Zeiss Distagon T* 2/35 @ f/2 (photo which really doesn't work as websize since the depth of field was too large, but as a panorama made with with 100mm @ f/2 or 85mm @ f/1.2 would make the subject really pop from background).
vahonen wrote:
Autopano seems to have Save As and import for Panotools. In theory this would give possibility to do all what is automatically possible in Autopano and then Save As Panotools format, then define the other control points in other software and save, then import the modified file to Autopano and finish the panorama.
I also tried to export to Hugin and add points manually. That didn't work since in this test I used camera-JPGs and they actually do not contain image date turned vertically, but instead one flag is set differently on header. It seems that Hugin doesn't understand this flag and images are 90 angle, and whole thing doesn't work.
Plan starts to form now, there are few things I need to test:
Offline
vahonen wrote:
Plan starts to form now, there are few things I need to test:
1. method: take picture with largest and smallest aperture of each frame, layers in Autopano, use smallest aperture pictures to create the control points for panorama. 2. method: take pictures as accurate spacing as possible, on same order as gigapan and use gigapan import in Autopano (extreme difficult to do manually with 180mm & 300mm lenses). 3. method: in addition to narrow depth of field images, take few with wideangle and huge depth of field, use Autopano and layers to create panorama. 4. method: import whatever method to Autopano, export panorama to Hugin, add control points manually and save, finish in Autopano.
Alexandre promised once to add a Replace button to the Add and Remove ones in the layer editor, every interested APP/APG user should help me to get it as soon as possible !Offline
Thanks for sharing wisdom, GURL!
Method 1 starts to look very good even not so nice to shoot. Maybe I have to study Canon's development libraries how difficult it would be to create program which takes two photos with preset shutter speeds, apertures and ISOs by pressing one button in computer, also would need to parametrize the mirror lock-up and delays between exposures and from mirror lock-up to actual exposure. Method get's too slow if I have to change from f/1.2, 1/4000s to f/16, 1/30s in each photo. Also risk to nudge camera position slightly between exposures increases.
Good to know that in .pano file I can just rename the images, I feel more confident to do that than let Autopano do "something automatic" with the layers/stacks.
GURL wrote:
# 4. I would bet APP being better at placing CP on fuzzy images than everybody doing that manually (I'm an experienced "manual CP gamer" who was impressed by the CPs APP was able to set in the above example!)
Maybe I placed the frames not as well as you - I did not get any points by automatic control point thing, only by adding hard link between blurred images I got 4 control points which were not so good (could have done better manually, but no possibility in Autopano).
Damn, could not wait to start experimenting these things, but have to wait two long weeks before I can start...
--
Samuli Vahonen
http://www.vahonen.com
Offline
Hi Vahonon,
what beauitful lens's you have!
Off the bat, I think if you connect your camera to your laptop (if you have one) I think Canon's DPP software, will let you make those controls, changing settings without touching the camera, I don't have the wireless remote kit, but I would think you could even use this one too.
Now, what fma38 and karl et al are working on, using the Merlin, this will give you more option as well as automate the whole procedure, at least that is my understanding of it, and therefor also my interest in that project.
but carrying all that gear around... ;-)
good luck
Henrik
Offline
Hello again,
I finally got the Nodal Ninja 5L and it seems that this is much less issue when using panorama head. As an attachment here is shallow depth of field panorama (3 row, 20 column - 85mm f/1.2). I had to place 7 images by hand but it was really easy since all images were nicely in rows and columns, just look the pitch and yaw from neighborhood images.
However I will to experience the techniques and get back to here when I have some results.
--
Samuli Vahonen
http://www.vahonen.com
Offline
Ehehe thats exactly what I tried doing - using APP to get a real shallow depth of field image of an object too big and too close to the camera for doing it with a normal lens (or a normal lens that I own :] ). I also use APP to stitch ~5:1 macro shots, and those of course have very shallow depth of field. Overcoming this problem isn't easy. At such magnifications small nudges to the lens result in big changes in the image, so focus-stacking works very poorly. Not only badly-aligned images, but also the solitary fact that changes in focus need to be so big in order to capture the depth of the whole subject that things move around in a parallax way (hard to explain, i hope you know what I mean). Programs like tufuse fail at coping with these shots, e.g. I tried stacking 5 images of a fly's head but the sharp/blurred areas were in such different places that the result was poor, I had to mask areas manually.
Anyways I can only hope that future APP versions will be able to focus stack more 'intelligently' than tufuse/enfuse and make macro panography easier.
Offline
Pages: 1
