 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
For my New-York panoramas (see post in Panorama Gallery), I had to create a specific process to adress the fact that there were always many people moving while I shooted bracketed (3 brackets) panoramas.
I had to use PTGUI (in demo mode) with its masking capability (and associated mask saving) in order to accurately pre-process my images (generation of png images with holes) before performing the final stitch with APG.
The new masking tool of APG (the red/green target) is absolutely useless for complex bracketed situations.
My dream is to have something similar to what I have been using in PTGUI.
Offline
I prefer using photomatix pro for ghost-removing and bracketing.
APG 3.0 ... did you use (1.) the left red/green or (2.) the right ones?
the right ones work on stacks the left ones on single image.... ;-)
you also could use ND Filter and using - depending on focal used (because of fov) - long exposures to eliminate ghosts like cars and persons... bracketing is possible too using ND filters ;-)
I also use the promote control device and let trigger the camera multiple times for each bracketed position. than using prhotomatix or enfuse gui to reduce these images to one image with all ghosts removed and than using the different brackets to generate the fused or tonemapped image using photomatix or using these images in APG directly as stacks...
Georg
Last edited by gkaefer (2012-12-16 21:33:40)
Offline
One thing you might try is shooting multiple brackets of the same shot. Should average out most moving object.
Offline
My issue was not to remove the moving objects : they are part of the scenes and they contribute to give life to the panoramas. My purpose was to make sure that a given set of pixels of the resulting panorama would be either 100 % or 0 % a moving object (car, pedestrian...) in order to avoid ghosting effect.
The only way I found to do this was to use the masking tool of PTGUI: it makes it possible to simultaneously view two adjoining images and create very accurate masks (down to pixel level). These masks can then be exported and used in photoshop to generate png files.
I tried to use the red/Green markers shown by Gkaefer, but with brackets and many people in the picture, it is useless.
Note: what I call moving object is not necessarily an object in fast movement, but an object that may be present at one place in Picture 1/Bracket 1 and not be at the same place in Picture 2 or in Picture 1/Bracket 2. Not managing this will cause ghosts in the rendered panorama.
All the panoramas in jmh.trp.free.fr/New-York have been processed using PTGUI as a masking tool. They all are stitched from 7 to 10 pictures and 3 brackets/picture.
My request is to have APG implementing such masking tool.
Offline
I'm still waiting for a masking tool similar to PTGUI as well. I've tried the green and red markers for removing myself and nodal ninja. They don't work reliably for me. I've gone back to using Photoshop for the masks.
Another feature I'm still waiting on is the ability to designate an image to be treated as the nadir.
I want that image to be warped/stretched as needed to match the rest of the pano without allowing it to adjust any of the images that were taken from the tripod.
I also want that image to only be used to fill in missing bits. I don't want it to cover anything from the other images.
Offline
Removing tripod from nadir is sooooooo easy with the markers, just make multiple nadir shots. Normally 3 but 4 when shadows occur
Offline
Hans, I use this method: http://www.kolor.com/forum/p49511-2009- … 511#p49511
and unfortunately, I have not been able to get good results with the markers.
Can you detail exactly how it works for you? Step by step?
Thanks!
Offline
Jermiah, not much to describe. Just run the preview in the editor and if at some places there is tripod left, go to the right photo and place a red marker on it. A thin one if you are working with single photos, a fat one if you are using stacks. Actually I always use the fat one as it also works on single photos ;-)
http://www.autopano.net/wiki-en/action/ … sking_tool
If that does not work, how many photos have the nadir insight?
Last edited by HansKeesom (2012-12-20 09:52:52)
Offline
HansKeesom wrote:
Jermiah, not much to describe. Just run the preview in the editor and if at some places there is tripod left, go to the right photo and place a red marker on it. A thin one if you are working with single photos, a fat one if you are using stacks. Actually I always use the fat one as it also works on single photos ;-)
http://www.autopano.net/wiki-en/action/ … sking_tool
If that does not work, how many photos have the nadir insight?
Hey Hans!
Let me add that this tool in most of the cases works fine! It works best if you not only mark what you want to get rid of - but also what you want to keep.
So when you have an object to rermove and in the next image a prt of background where this object is in the other image - placing a green mark on the background to keep
helps making the removal almost perfect usually.
That´s quite a different way from the masking-tool in PTGui - but it works very good nevertheless when you get used to it.
best, Klaus
Offline
Hi Klaus,
Thanks for the tips, I will remember them and try to test them.. I not often have to deal with moving objects, more with small breaks in lines due to distortion or imperfect shooting of the photographer. Almost always I can correct it with the help of my red and greens friends ;-)
Offline
To Klaus:
I experimented the red and green marking with success only once: there was not bracketing and a very limited number of objects to remove.
But with bracketing and crowdy situations, it is awful and does not work. Furthermore, in those crowdy situations, you may have to be very accurate (down to pixel level sometime) and, without a tool like the one of ptgui, the situation is so complex with lot of superimpositions on the preview that you are unable to guess what the objects to keep or remove are.
Before using ptgui, I did it by iterations but it was not deterministic and time consuming. The ptgui approach is also time consuming but perfectly deterministic and the outcome is certain. If the set of images is such that there is no combination where you will be able to get completely rid of the ghosts you will understand it using the PTGUI approach. With APG only, you may iterate for ever without knowing for sure that there is no solution to your problem.
So, I definitively need something similar in APG.
Offline
