https://drive.google.com/drive/folders/1ds5QHI4qNYqXOLNexS0Ov6Ll14HnhkSV?usp=sharing

Hopefully you can see those files.   I retract my advice to use darktable's 
retouch tool -- there is a lot of sensor dust.

I ended up scaling the image by 50% because it was taking so long, but 
skyfill did happily work on the original full sized image.

I used the full sky replacement option (-fsr), which basically examines 
every pixel in the detected sky and assigns a probability that the pixel 
should be considered true sky.  If the probability is large enough it will 
be overwritten -- so magically all the sensor dust disappears.   That one 
spot near the horizon on the right where there is a black mask leaves some 
artifacts, because of some feathering at the edge of the mask where pixels 
were not completely black.

I will better document the -fsr option in the next tutorial for skyfill.

There is a shell script which shows the flags I used (just -fsr and a 
number of -tm flags).  I had to mask a few of the sensor dust spots because 
they were too severe and caused end of sky detection to stop at them.  The 
-tm mask tells skyfill to ignore that area for testing to see if the end of 
sky is there, and to not use that area as sample data to build the HSV sky 
model.

There are two runs of skyfill in the script, one with the -d3 option (debug 
level 3), which creates an image that shows the start of sky detected, end 
of sky detected, and where the masks are at.

A couple of other things:

Notice how that brown smear in the sky near the black-mask at the right is 
gone -- full sky replacement fixed that too.

The original full sized image caused a real problem to show up -- the large 
vertical shift in the sky data about at about 60% of the image width.   The 
problem is (and has been the biggest problem for skyfill), is pixels near 
the edges of the data can be just "wrong".  I'm not sure where the problem 
comes from, but in this example it is because the pixels on the edge near 
x=16278,y=551 are too dark.  Skyfill in the first few steps will fill in 
pixels just to the left of that vertical edge because it is not perfectly 
vertical, and is using a localized rgb model to extrapolate and create 
pixels to make the edge perfectly vertical.  Those dark pixels pollute the 
localized estimate of sky color...

On Friday, March 4, 2022 at 4:41:32 PM UTC-8 Jeff Welty wrote:

> I'll grab that, woah, yes that's large.  As an afterthought -- if you have 
> darktable -- just use the retouch tool.  It'll be quick, easy, and like 
> magic!
>
> On Friday, March 4, 2022 at 4:25:22 PM UTC-8 [email protected] wrote:
>
>> My existing examples are absurdly large.  I don't mind putting one on a 
>> google drive freely accessible.  But you might mind trying to download and 
>> work with it.
>>
>> For several reasons, I prefer to fix the gaps after blending the 
>> panorama, rather than fix individual photos before.  One reason is that 
>> gaps might go away during blending because images overlap.  In that case I 
>> would prefer to end up with original from the image that doesn't have a gap 
>> there, rather than synthetic from the image that does.
>>
>> I put a 680MB image on a google drive in case you want to look.  I'll try 
>> to make something more reasonable within the next few days.  If you look at 
>> that, there are three tiny gaps that I think are transparent (look white in 
>> a viewer) plus one made exactly the same way but looks partially white and 
>> mostly black.  I have to figure out why that one looks that way (unless 
>> your tool just fixes it).
>>
>> https://drive.google.com/file/d/1eLvJsyDOrV06bc1YLAQRoG-1x4vv5gb4/view?usp=sharing
>>
>> The whole panorama (taken in valley of fire NV) is still a work in 
>> progress.  There are a few blurred seams I want to figure out how to fix.  
>> So I'm looking for how-to for this kind of thing, rather than fix this one.
>>
>

-- 
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
--- 
You received this message because you are subscribed to the Google Groups 
"hugin and other free panoramic software" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/hugin-ptx/137e360c-df63-4f01-8979-52df124475f2n%40googlegroups.com.

Reply via email to