Indeed, small misalignments can cause blur as too, but the blurryness
would be limited to a small band (the blender uses a small transition
band for high frequency content). This doesn't seem to be the case in
Richard's example.
Joost
>
> Hi Roger,
>
> Indeed, small misalignments can cause blur as too, but the blurryness
> would be limited to a small band (the blender uses a small transition
> band for high frequency content). This doesn't seem to be the case in
> Richard's example.
Hmmm. In that case I am probably experiencing the same problem as Richard,
without knowing it, because the blurring I experienced was not confined to
such a narrow area. I remember painting large areas of the bottom of
several panoramas. Come to think of it, one of them had a gravel base,
so there was no wind ripple!
What else could it be?
Roger W.
Can you borrow another camera from somewhere ?
So you can isolate that as a cause ?
regards
mick
> I'll try some more exteriors tomorrow shooting just four-around. But
> I'm still bewildered. For a while I was convinced I had a dud lens,
> now I'm thinking it's maybe exterior chromatic aberation effects on
> stitching and poor angular separation ?
Chromatic aberration is a killer. Any feature ocurring on the left of
one image and the right of the next will have its coloured fringes
on the opposite sides! This makes it very difficult to stitch well.
It also makes traceries of leaves and branches against the sky, often
occuring near the zenith, of course (and therefore very near the
edge of the field of view where CA is strongest) look quite fuzzy.
Stitching after having removed CA is FAR superior both in ease and
quality of results. Well worth doing, in other words!
I have read your email to the fpp list regarding cube faces and how to
produce them using PTGUI, thanks you very much for that.
In your opinion which is the best size for a Maximum size ("no loss of
detail")panorama of say 11000 x 5500 and what % quality will you use?
kind regards,
Rodrigo
Richard schrieb:
> I'm using a Sigma 8mm 3.5 with a Canon Rebel XS & PTGui 8.1.4. I
> usually shoot 8 around at 5-degrees up.
>
...just an idea...
Aren't 8 shots with a Sigma 8mm overkill in therms of overlaping?
Probably some pictures arent used for the final panorama and/or
the seam badly positioned. The NPP is also varying with different
pan angles.
Bes regards
Peter
> CA is a fixed thing for every lens...
To be correct only transverse CA (TCA) is fixed, longitudinal CA is
reduced by stopping down.
TCA will most likely vary with focal length of a zoom lens. And it might
change with focusing distance even for prime lenses. More information on
http://wiki.panotools.org/TCA
--
Erik Krause
http://www.erik-krause.de
And in addition, fisheye lenses TCA varies with the light wavelength
(e.g. incandescent / sunlight / fluorescent / etc.):
http://tinyurl.com/32jjdz
Sorry, unlike all the rest, the Sigma 8 mm f3.5 lens was tested here
under a single lighting.
The blue channel is most affected and especially at or near the edge
of the coverage. In contrast, the red channel seems not to spread too
much from the (reference) green channel.
Hence a generic correction strategy can generally be adopted
satisfactorily. But specific TCA correction settings should be
applied for getting the **best** output... when you look at a corner
of your source RAW at
200% (as Joergen has suggested;-)
Regards,
Michel Thoby
> And in addition, fisheye lenses TCA varies with the light wavelength
> (e.g. incandescent / sunlight / fluorescent / etc.):
> http://tinyurl.com/32jjdz
I suspect this is at least partly an effect of longitudinal CA. CA is a
property of all optical media and clearly wavelength dependent. CA is
BTW the reason why there is a rainbow or why a prism generates a spectrum.
Lenses are usually corrected for LCA, but there are different types of
correction and it never is perfect. Depending on how much wavelengths
are corrected there are f.e. achromatic (2 wavelengths), apochromatic
(3) or superachromatic (4) lenses. There are nice diagrams on
http://toothwalker.org/optics/chromatic.html (scroll down to Figure 5 -
"chromatic" means a totally uncorrected lens, like f.e. a loupe)
This shows that an achromatic lens has most LCA (and probably TCA, too)
left in the blue channel. If you compare channels at 200% or higher, the
blue one often is the most blurred one. This is because the correction
curve leaves the zero-line progressively with shorter wavelength.
Now if you have different types of light, they mostly differ in the blue
channel. Incandescent light spectrum blue side ends at longer wavelength
than sun spectrum. Fluorescent light has no contiguous spectrum at all
but consists of several narrow bands. Could well be only one such band
is inside the range of the blue filter. If two bands are inside, they
could well cause what you call ghosts in http://tinyurl.com/2zzjsl
If the shortest wavelengths from the blue channel are missing, it is
pretty clear that both chromatic abberations are smaller, since the lens
might be corrected well for the long wavelength end of the blue channel,
but not for the short wavelength end.
Reading your last post, something crossed my mind: would it be better for the overall IQ of the images- to only select the blue channel to correct the CA, and maybe after that, resellect all the channels and give a final correction of the CA, if needed ?
;-)
Am 26.04.2016 um 13:09 schrieb hamid.m...@gmail.com:
> I have the exact same problem. I am using a Canon 60D and Sigma 8mm 3.5.
> Did you ever find the reason for the blurriness and could you resolve it.
You revived a very, very old thread. The examples in the quoted message
are no longer available. If you make your own examples available, may be
we can help.
Thank you for uploading those sample panoramas. The problems at the seams at the nadir are very evident, and the uncorrected chromatic aberration is quite bad. But we really need to have a sample set of camera images. I hope that you shoot raw and that you can supply copies of just the 0ev exposure set. If no raw shots are available, then medium quality jpegs will be fine.
I uploaded the camera images (12 images) as well as a new panorama I created that is stitched better (not that it makes a difference to main blurriness issue).I need to process these images in batches so correcting them individually is not really an optimal solution.
John
That actually looks really good. I guess I could do the same to the merged HDR images and then run them in PTgui. How did you correct the chromatic aberration? I tried and did not get a match better result.
Hamid, M mode is strongly recommended as the blending is so much easier when the images match in the overlap areas. Don't use auto white balance either. Blurring will be no different, except that you will not have degradation of the image quality due to the blending of images having very different brightness levels, as can happen in Av mode.
I also tried removing colour fringes additionally to the automatic chromatic aberration correction (Adobe Lightroom).But the items, walls, etc with same colour in the fringe range were desaturated, so they look grey/strange.I tried to narrow the colour range down, but this also made the fringe correction useless.This is the reason why I currently only use the automatic chromatic abberation feature - it works pretty well except in extreme cases (the light brackets) -but I cannot go through each bracket shot and intensively check for desaturation - or even mask them out manually, it would take far too much time.Maybe there is a better tool or way for this?
More contrast gives me more blur. Eg. When dialling exposure right down to get the windows in the church right, (making everything in the image pretty much black) I get more blur.
Probably is a setting somewhere. Advice on the perfect computer workflow much appreciated here.
Richard
Yes. Full manual. Using one raw image and sailing down exposure 2 stops, normal and then up 2 stops. I had a play with TIFF as was mentioned and get an absolutely perfect stitch with no errors in the floor detail, and no blur, but this is without any blending of other layers. When using tiff with blending I still get the blur.
uploading 8 images now https://www.dropbox.com/sh/q7m30zy0ufucaeh/AABvecQXz-63Ga01cDfah8Vba?dl=0
Ah, sometimes I've had the know a bit loose to make the rotation easier - I'll make sure it's tighter. I'm thinking this isn't going to affect parallax at 20 odd feet away...
Had a play with it yesterday and the tightness, when loose to allow easy rotation, causes the NN to fall off the vertical by quite a way. I didn't realise before that this was affecting parallax.Nonetheless, it still doesn't explain why I've had this blurryness in RAW converted files and not when I've blended JPEGs.