Page 1 of 1

That old and recurring autopano bug

PostPosted: Mon Sep 24, 2012 11:58 am
by hermer-blr
Over the week end I have stitched (APG 3. Beta - Windows 32) a panorama with a lot of sky:

- about 10 images horizontally aligned (X 3 for bracketing) shot with telelens
- color correction set to LDR
- multiband level set to 0

The town below the sky is perfect but the sky at, say, 5 different places over the width of the panorama is unclean, grey (the usual bug). In the past, I have seen several posts about this bug. What is the present situation in term of bug resolution ? What must I do to improve the sky ?

The preview with other multiband levels, either negative or positive, is worse than with 0...


PostPosted: Mon Sep 24, 2012 4:58 pm
by DrSlony
Can you please show a screenshot of both the rendered image and of the pano editor window with ayers and coor anchors visible?

PostPosted: Mon Sep 24, 2012 6:51 pm
by hermer-blr
Here are some screenshots and the upper left part of the panorama showing a dramatic example of the problem.

Thanks for your help

PostPosted: Mon Sep 24, 2012 8:06 pm
by klausesser
hermer-blr wrote:Here are some screenshots and the upper left part of the panorama showing a dramatic example of the problem.

Thanks for your help


Try this:

deactivate "optimize by layers"
set the reference anchor to a sky-area
use only "exposure" with the anchors.

best, Klaus

PostPosted: Wed Sep 26, 2012 12:08 pm
by hermer-blr
Thank you for this advice, Klaus.

I have done as you propose and will get the result tonight, when I am back home. What I however noticed is that the preview has remained unchanged (grey areas still there) following these modifications. I did not change the position of the reference anchorS (1 per bracket), as all the images (1 raw) exhibit the same proportion of sky.

Best regards

PostPosted: Wed Sep 26, 2012 12:28 pm
by lumelix
If your test isn't successfullym try once again whithout any reference images/anchor points.
I try out and test the new 3.0 for color correction and found, that without ref/anchor, the results are the best in blending.
Even when overlapping zones are handmade and irregular like in your case.
Have you try to correct the vignetting before stitching with APG? It looks like there is a lot of this.

PostPosted: Wed Sep 26, 2012 8:15 pm
by hermer-blr
Klaus solution : no change

Lumelix solution : I am testing it but the preview does not show any improvement... I will report the result when the processing is over.

And Kolor, what do they say about it ? This bug has often been reported in the past. Is it supposed to be corrected, or am I wasting my time trying to find a solution while there is not ?

PostPosted: Wed Sep 26, 2012 8:52 pm
by lumelix
If it's the blending/color correction itself, Kolor says this is a priority for V3.1.
But what you think about vignetting. You only have one row and the dark zones are
always between the images. Until now, APG did not correct vignetting.

Btw. For tests, render at 50% or smaller. Faster and you see if it's ok or not.

PostPosted: Wed Sep 26, 2012 9:12 pm
by hermer-blr
To Lumelix: I confirm that the rendering has not improved with your suggested solution; regarding vignetting, I have not tested it. What method would you suggest to process it ? I am not sure that Nikon Capture NX2 can perform an automatic vignetting correction. It will only apply the value that I will set up : not optimal at all !

PostPosted: Wed Sep 26, 2012 10:58 pm
by gkaefer
for automatic vignetting correction...if your lens and nikon are supported (if profiles exist), I would suggest:


PostPosted: Thu Sep 27, 2012 10:36 am
by lumelix
Hi hermer
I use NX2 for correct vignetting, also when shooting panos with the AFS300mm 2.8 or with TC20E.
But it's not automatic. I have to find the correct value for the usual used aperture. Most of time i shoot
wide open (2.8 / 5.6). The correction value in NX2 is then around 50%.
You can try out the correct value by shot a piece of blue sky, copy a small section in the corner to the center and look
if it differs. If not, you have found the correct value for this settings.
Once you have found these settings, you can store them for later recall.

I don't know if there is a software that can do this absolutely automatic for every camera/lens/converter/aperture combination.
But I suspect not. Adobes PS and its lens profiles doesn't do the job as well as NX2 (and I found that RAW conversion is also better in NX2).
Maybe DxO Optics Pro can do that, because this guys are testing a lot equipment. But I haven't test it until now:

PostPosted: Thu Sep 27, 2012 11:45 am
by hermer-blr
Thank you, Lumelix.

I will give a try to your different suggestions. However, I suspect that vignetting is not the cause of the problem. Would APG not correct the vignetting, not a single panorama would be acceptable.

The software has for long had an issue with the sky; I thought that it had been resolved, but it seems not...

The question is now are there suggestions that would help in a not to painful manner to stitch the panoramas without these grey areas in the sky...

PostPosted: Thu Sep 27, 2012 12:38 pm
by klausesser
hermer-blr wrote:The question is now are there suggestions that would help in a not to painful manner to stitch the panoramas without these grey areas in the sky...

The "grey areas in the sky" might be related to the use of Gamma-anchors. Try using only Exposure-anchors. In my experiences the grey-ish
areas do not appear then.

But i agree basically: the color-correction still has it´s issues as it seems.

best, Klaus

PostPosted: Thu Sep 27, 2012 3:35 pm
by lumelix
The reason why I think about vignetting is because I never have such an issue with a single row pano.
And I found that the color correction since APG V3 A2 is really improved.
But as Klaus wrote, the color correction in APG have still some issues. I found a lot of errors when using gamma
and exposure correction together or with color correction in giga panos with more than one row.
You have to note that APG don't correct the vignetting until now. It does a little optimization while blending,
so vignetting is less visible, but this can not replace a real vignetting tool.
I found in giga panos, it's really necessary to correct vignetting first, even with the 600mm combo.
Otherwise, vignetting is always visible as darker bands between the images, most in the sky.

Try Klaus's suggestion and use once only exposure correction with no or only one "reference image".
You can also completely deactivate the color correction to see what happens. If you have still these grey areas,
it has nothing to do with the color correction.
Try also different multiband levels form -3 to 0 and try other than multiband (nothing or linear) to see what happens.
For tests you can render the pano really small, 20% or 50% or so. Then you have the result faster.

I just see that you have dehazed the images. And when I'm looking to the clouds, it looks something like tonemapped ?

PostPosted: Thu Sep 27, 2012 6:13 pm
by hermer-blr
Thanks to both of you for all these suggestions.

I have tested as suggested by Klaus : exposure instead of gamma - No improvement.

I will try with no color correction... but does that make sense with 3 exposures and exposure fusion ?

I have not tonemapped the images: I have shot 3 brackets and put all the images in Autopano, to perform exposure fusion, as I use to do. I have dehazed (50%) the under and over exposed images; I have also made a try with all 3 exposures dehazed.

PostPosted: Thu Sep 27, 2012 7:37 pm
by lumelix
Oh, I have overlook this in the first post. Have you done bracketing by exposure time? Certainly!
I'm not shure but with braketing you should better use the HDR option in color correction.
Workflow case E. Did you so?
What is if you take only one of the exposures, eg. the one where the sky is well exposed?

PostPosted: Thu Oct 04, 2012 5:55 pm
by hermer-blr
I would be interested to have a reply from KOLOR, at least a confirmation of the bug, if possible an indication of a sensible workflow to process these images...


PostPosted: Fri Oct 05, 2012 4:19 pm
by ThomasV

It looks like a vignetting issue. For more details on that, you can give a look at

It is for us a very important issue, and therefore, we plan to work on color correction in Autopano 3.1.


PostPosted: Fri Oct 05, 2012 5:32 pm
by hermer-blr
Thank You - This is a long known bug, that was already there in 2.X - I hope that it will now for sure be corrected in 3.1

PostPosted: Sat Oct 06, 2012 9:31 am
by hermer-blr
I just downloaded DXO optics pro and used it to process my raw files, including vignetting processing.

The APG result is now perfect, even in the Gamma correction mode.

I must admit that I could not believe that APG was not able to perfectly correct vignetting : it should be part of its basic features.

As I understand it, DXO optics pro has a database of cameras and lenses. It sure helps.

Thank to all for your help and suggestions

PostPosted: Sat Oct 06, 2012 10:15 am
by leifs
I have been searching for the perfect workflow and tried several. For now I'm happy with this flow:
- shooting raw spheres with the E-M5 and Leica 25mm f1.4 @f8
- developing with DXO - which has modules for both the house and the lens: remove vignetting and lens distortions
- detecting in APG with "Lens distortion correction" disabled

This gives me a much better result than e.g. developing in Lightroom 4 and use the APG "Lens distortion correction"
In DXO it is important that there is modules for the lens/house combo.


PostPosted: Sat Oct 06, 2012 4:43 pm
by hermer-blr
Thanks for this advice, Leifs

In my own case (Nikon camera and Lens), DOP had the right information in its database.

But I have not disabled lens distortion correction...

The panorama looks however perfect... at first glance. However, when I darken the sky (e.g. with a progressive filter in Nikon Capture NX2), some greyish areas tend to reappear. So, there is still an issue that lens distortion correction disabling might have permitted to avoid...