lumelix wrote:

Let me know when you found a solution for your problem.

Hi Martin!

Thanks for your fast reply. Yes - it´s in the rendering also. I did the sreenshot without cc on - there´s no big differences. What´s funny: watching the screen when cc is off there are changes nevertheless - constantly changing luminance per image while the wheel showing texture-mapping runs. It takes about half an hour anbd the the image has changed. But not for better . . .

When i switch cc to "Auto" at the start it looks quite good - but the the "hanging-process" starts and in the end it looks like in the screenshot.

I wonder whether it is related to the fact that the sky-images are not linked? They´re placed by the xml - but have no control points. Is there an issue with color-blending in that case?

best, Klaus

P.S.: i´ll post another screenshot of the editor in the afternoon!

P.S.2: let me add i never had this issue before! I´ll re-start from scratch without any correction - i shot in manual mode anyway. But i realize that the single images in the editor change without having cc activated!
Could it be related to the fact they´re RAWs?

Last edited by klausesser (2012-03-18 13:55:12)

Hi Klaus
In my case, all images have CP's and are linked. But I have too this blending issues.
The images have same manual exposure and all EXIF's are ok. No vignetting or other problems.
I work then with JPG's because of the great amount of data.

I do not quite understand this funny effect you described above. It's normal that the images are flickering in the editor window when you are zooming or moving around and the render engine is working. But for a few images like this it should not last half an hour.

If it is true that the CC optimization does not take into account images without control points, then this is a big problem in APG.

Some tests:
Try CC only with exposure mode, no gamma or color correction. Set this to all images.
Check the anchor points for CC. Try first with one AP, then set some more, eg. a AP every 10 image
Try to convert the RAW to JPG. But preserve EXIF.

Hi Alexandre
The sky in the middle of the pano looks like there could be vignetting (asymetric, more at the top of every image, so the sunshade shouldn't mount properly .
But when Klaus says there is no vignetting issue with this images, I believe him.

But now look at the copper green building: There are areas over several images, that are much darker or lighter than around. So this looks very similar to the blending issues I have posted last year.

I don't know the reason for this issue. If there is something in the EXIF information ? Or it is more likely a calculation error from the blender funktion.
But the result is definitely a useless panorama !

I think both Klaus and I are quite capable to capture images for panoramas without any error. So for me it is definitely a malfunction of APG blender.
Please attempt to finally get this problem under control.

lumelix wrote:

If there is something in the EXIF information ?

Hi Martin!

Maybe - there are no EXIFs for the lens because it´s a Nikon adapted on a Canon 5D2.

I´d like to know what happens when images are well positioned but are not linked! Obviously they can´t be blended correctly BECAUSE they´re not linked together. This is what i read from the patern.

best to you, Klaus

lumelix wrote:

So for me it's clear that there must be a bug and big problem in the blender engine of APG, that produced this blending issues in panos with more than only a handfull images.

I´m afraid so. Here´s an example with Multiblend at "0" as Vincèn suggested. You can see that the falloff comes fron the center. So it can´t be vignetting - this is symmetrically.
In this case the falloff in each image has a direction.

I wonder whether SmartBlend might help here . . . !?

best, Klaus

Last edited by klausesser (2012-03-19 12:58:57)

lumelix wrote:

So perhaps it is not a vignetting but ghosting problem ?

I checked the lens and made test-shots: no vignetting at all. Besides: IF there would be vignetting the colorcorrection is there to compensate it!
The colorcorrection usually works very fine and equalizes areas perfectly. So i´m sure it´s related to the missing links between xml-positioned images
which Multiblend or the colorcorrection may have issues with.

Some more detailed information about what setting  to use in which case woud be really fine . .

best, Klaus

lumelix wrote:

Hi Alexandre
The sky in the middle of the pano looks like there could be vignetting (asymetric, more at the top of every image, so the sunshade shouldn't mount properly .
But when Klaus says there is no vignetting issue with this images, I believe him.

But now look at the copper green building: There are areas over several images, that are much darker or lighter than around. So this looks very similar to the blending issues I have posted last year.

I don't know the reason for this issue. If there is something in the EXIF information ? Or it is more likely a calculation error from the blender funktion. But the result is definitely a useless panorama !

I think both Klaus and I are quite capable to capture images for panoramas without any error. So for me it is definitely a malfunction of APG blender. Please attempt to finally get this problem under control.

Sure ! Let's distinguish 2 issues ( the first one posted as a screenshot in this first message of this thread ) : here, there is some issue in color correction were the exposure of individual images cannot be calculated. We are aware of this issue but it needs fix that are bigger than a simple fix. It will come only in a big release, not into a small one.
So for this case, as everything is being well shot, we don't have any exposure differences ( Klaus did it with manual exposure ). So no color correction needed.

Then the second issue : why these geometrical lines in the sky ?

My first guess is vignetting. As in the quoted other case ( http://www.kolor.com/forum/viewtopic.ph … 308#p90308 ), it looks like the same.
This is a pure vignetting case and the result after the blending : http://www.kolor.com/forum/p90205-2011- … -31#p90205
Looks like the same, don't you agree ?
Now how to prove that and how to compensate that. We'll need at least some of the input image to check that here.

Could it be in the blending engine ?

We don't do any radius calculation to would be needed to compensate for vignetting. All color operation on the pixel of individual images are global operator ( like histogram ). So how could we introduce such an effects that is spatial without doing spatial operation ? We could raise the effect like a histogram would raise the vignetting, but we could not create it.
So definitively, Klaus, we need some of the input images.

AlexandreJ wrote:

klausesser wrote:

I checked the lens and made test-shots: no vignetting at all. Besides: IF there would be vignetting the colorcorrection is there to compensate it!

Wrong. We don't do vignetting correction in color correction and we never did ( it's still on the todo list ).

Hi Alex!

I didn´t mean vignetting correction in special but equalizing differing areas in general - so that vignetting shoiuld be equalized also. Wrong?

best, Klaus

Hi Alexandre
Thank you for your hint.
I'm happy to see that multiband level 0 helps to avoid this first problem with different exposure in some area like this shaded wall on the left side.
It is clear that with level 0, 100% of image area is used to calculate the corrections and this is better in this case.
The only reason why I never try level 0 was, because I read somewhere in the forum not to use level 0
So I will try it with my problematic pano !

The second issue, I guess, have something to do with ghosting. Klaus used a big f1.4 85mm lens, so the sun can shine on the front element of the lens (in several images) , and this occurs in ghosting.
If you look closely to the image from Klaus, you can see some gradient color movement. But it is far better on the opposite to the sun.
This is typical for ghosting effects. But it is not the same as vignetting.
Vignetting can be corrected by a radial calculation. Various apps can do this, eg. Photoshop or NX2. And APG in the future
But ghosting is an other thing ! I have this too and it is very hard or near impossible to correct the effect of ghosting between two images.

But we can only be sure, if Klaus check this in his images by compare the center part with the edge in some of the affected images.
A side by side comparsion  of several images will be usefull too. Most of time, ghosting can easily be checked this way.

Last edited by lumelix (2012-03-19 14:34:00)

this is the best one i could achieve:

AlexandreJ wrote:

Klaus, could we have this check to find out about ghosting versus vignetting. Do the following rendering :
- blending mode : linear
- uncheck ghost
- check 'diamond weighting'
Render and compare to the previous rendering

I´m back right now and do the rendering - coming son . .

best, KLaus


here it is:

I i put 5 sky-images on my server: www.klausesser.de/kolor_upload.zip

and again 9 images frrom the same scene: www.klausesser.de/kolor2.zip (uploading right now)

Last edited by klausesser (2012-03-19 18:39:41)

Hi Klaus
There is a lot to discuss
But first, there is something like gosting in this images (5 images in first download, second one won't function for me).
Have a look at the two images below:
In the first you can see that there is a big difference between every image. They have all a diametral color and brightness shift from bottom right to upper left corner.
The difference in brightness goes from 47% to 56%. This 9% equals to 1EV, when using the 5D MK2 (about 11EV dynamic).
In the second image I have push the contrast to it's optimum. As you can see there is too a slight radial component and a color shift.

I have no idea where this is coming from. It could not be a simple ghosting from sun, because it is in the images too, where the sun is standing behind you.
Perhaps it is from the light floor. Or it goes through the viewfinder.
Or do you use a shift lens ??
In EXIF data I found that the lens is 50mm (not 85mm) but I couldn't found the aperture. This could confuse the CC if used.

Then I see that there are no CP between the upper and the lower half of the panorama. What have you done for this ? It should be no problem to detect CP's there !
Then the image line is very wavy, it should be straighten.
The rms is very good (near 1), don't worry about this.

In my experience it is almost impossible to correct such "ghosting" effects completely, even manually in PS with only a pair of images.
Therefore, you should look for the reason of this effect and try to avoid it.

Last edited by lumelix (2012-03-20 01:40:50)