 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
I usually shoot with fisheye lenses but I wanted to experiment with non-fisheye and multirow.
I shot this pano using my Nikon P5100 compact with the Nikon WC-E63 (0.63x) wide angle convertor which provides a rectilinear wide angle combo of approx. 24mm on my NN5 head using the T-30 adapter.
I'm reasonably confident that NPP is set correctly.
I shot 3 rows of 10 shots - at zero and +/-40 degrees.
Yes, I know this doesn't give a full 360x180 coverage and leaves small holes at zenith and nadir but that's fine with me.
I think the overlaps are sufficient but I cannot achieve a good stitch and my attempts at improving link quality through CP editing just seem to make it worse.
Here's the image set if anyone is willing to show/tell me what I'm doing wrong - 30x5MP images shot on manual settings 1/4 at f4.8 and ISO 100 - be warned it's a 50MB ZIP file!:
http://www.three60views.org.uk/image_se … 63_set.ZIP
Last edited by mediavets (2008-09-09 11:48:45)
Offline
Okay, this case is really interesting as it illustrates well one fairly unknown issue with lenses : focal stability.
In fact, exif datas are showing exactly the same value, but there are not exactly the same in real.
You can check that in a 2 passes stitch :
- 1st test : just use the images as provided. The software will check the exif and as the focal is the same, it will optimize only for one focal for all images.
- 2nd test : open the image properties of the group. Select just one image and put a really different value ( I've put 30mm instead of 23 mm ). This will force the software to optimize all images with their own focal. And the result is far better in fact. This shows that the focal wasn't really constant during the shoting.
Offline
AlexandreJ wrote:
Okay, this case is really interesting as it illustrates well one fairly unknown issue with lenses : focal stability.
In fact, exif datas are showing exactly the same value, but there are not exactly the same in real.
You can check that in a 2 passes stitch :
- 1st test : just use the images as provided. The software will check the exif and as the focal is the same, it will optimize only for one focal for all images.
- 2nd test : open the image properties of the group. Select just one image and put a really different value ( I've put 30mm instead of 23 mm ). This will force the software to optimize all images with their own focal. And the result is far better in fact. This shows that the focal wasn't really constant during the shoting.
Thank you Alexandre. I have learned something new about cameras and about APP yet again.
Is this 'focal instability' (for want of a better term) common ? More typical of compact cameras than DSLRs? I guess it's just a mechanical issue with lens focusiing and zoom mechanisms?
And this set of screen shots the difference following your suggestion:
First image shows result using EXIF data, quality of detection setting Standard, 50 points/image pair.
Second image shows result using setting of 30mm for one image in the set, quality of detection setting Standard, 50 points/image pair.
Third image shows result using setting of 30mm for one image in the set plus manual CP editing to add and finesse links, quality of detection setting Standard, 50 points/image pair.
Fourth image shows result using setting of 30mm for one image in the set, quality of detection setting High, no further CP edting, 50 points/image pair. This produced the best stitch.
Forcing APP to compute the focal length for each image - and optimise accordingly - resulted in computed focal lengths between 20.96mm and 22.14mm. The EXIF showed 23mm.
Last edited by mediavets (2008-09-10 14:27:37)
Offline
This one owes a special page on the wiki directly.
Offline
So if there is even one exif focal length that differs from the rest, then APP will ignore those values and work them out itself.
When, in what situations, will using this exif focal length give us best results?
When, in what situations, will forcing APP to detect the focal length itself give better results?
Offline
DrSlony wrote:
So if there is even one exif focal length that differs from the rest, then APP will ignore those values and work them out itself.
There are 2 modes of focals calculation :
- all image have the same focal and lens distortion
- all image have individual parameters ( focal and lens distortion ).
The rules to choose between both mode is calculated from EXIF data : if all focal are the same, the first mode is used.
By doing the trick exposed before, I break this rule and force APP to use the second mode.
DrSlony wrote:
When, in what situations, will using this exif focal length give us best results?
I seldom saw such a behavior with DSLR ( except with zoom lens where the operator had slightly moved the zoom while making focus ).
Those camera are really stable. This case is more typical of compact with convertor.
DrSlony wrote:
When, in what situations, will forcing APP to detect the focal length itself give better results?
Don't make me wrong, there are not a lot of case where forcing APP to calculate all focal length is better than the other mode. The other mode is first really faster because you have 3 unknowns to calculate instead of 3 unknowns / image for the second mode !
If you are not sure of the focal, try this mode ( BTW : it's the one used for images without exif ).
Offline
Hi Alexandre,
If the focal length does not change, but the NPP does (e.g. I move the camera backwards or forwards) does APP se that as a different focal length, or just a change in the 3 unknowns?
BTW, what are the 3 unknowns? If its X, Y, and Z (e.g. milli- or micro-meters from the center point), then don't you also need pitch, yaw and roll?
Offline
