Am 14.03.21 um 12:04 schrieb Harry van der Wolf:


Op za 13 mrt. 2021 om 17:20 schreef 'Kay F. Jahnke' via hugin and other
I also used the new --snapshot_threads and set it to 3.
However, pv still uses all 4 cores/threads and when I "time" the command it is exactly the same as without this parameter.

Sorry, that was a quick shot: batch jobs always set the number of threads to the number of cores, the option --snapshot_threads only affects the number of threads used when the job is launched via the UI with 'U' or 'Shift+U'. I'll change it so that the option is honoured always.

    I don't look at the
    masks in the PTO yet, but you should now be able to simply feed images
    with unwanted content 'erased' to transparency (should even be best
    with
    a feathered brush on the erase tool), and as long as the other images
    provide content, the 'ghosts' should become invisible, thanks to the
    B&A
    magic.


I do not entirely understand what you mean. Do you mean I should erase some unwanted part in one of my bracketed??images? or does pv that? Or do you have to do that in the live view?

Let me explain in detail.

Sometimes you have things in your images you don't want to see in the final output. In a stitching/image fusion context these unwanted bits are often called 'ghosts', and the process of suppressing them so that they aren't visible is called 'deghosting'. For stitching, this is often done automatically, and a commonly used method is 'Khan deghosting', used e.g. by hugin, see http://hugin.sourceforge.net/docs/html/namespacedeghosting.html

pv does not have automatic deghosting (yet), so if an image of the set you are fusing/stitching has unwanted bits, you must mask it out 'manually'.

In hugin, you can assign an 'exclusion mask', which results in the partial image, rendered by nona, coming out transparent where the mask is set. But pv does not process hugin's mask information (yet), so if you want parts masked out for processing with pv, you have to 'manually' make the unwanted bits transparent. You can do that e.g. with gimp: add an alpha channel to your image if it doesn't have one yet, then pick the 'eraser' tool and select a *soft* brush to 'erase' the unwanted bits with a bit of 'feathering'. Store the image. If pv reads such an image (it must be run in alpha mode, so use pv --alpha=yes) it will exclude the transparent parts - but not totally, instead it will honour the feathering.

This is really interesting. I personally like sunrises and sunsets. You mostly need bracketed sets to get a good dynamic range. They sometimes contain unwanted parts or enfuse (up til now my used program) does not pick the??part I want. I sometimes fall back on hugin for its use of masks. (enfuse does understand masks, but I do have to create them in hugin so I can just as well finish in hugin)

You're right, enfuse does understand masks, in fact I think enfuse expects all it's input to come with an alpha channel. How you create the transparency - hugin mask, then nona - or e.g. the gimp using an eraser - is irrelevant. But enfuse has a limitation (to my knowledge, please doublecheck): it's either-or for enfuse: if there is the slightest bit of transparency in an pixel in an incoming image, it will completely exclude the pixel. And hugin also has a limitation: the masks are either/or as well: inside the mask is totally transparent, outside totally opaque.

pv, on the other hand, will accept both images with alpha channel and images without alpha channel. And if there is an alpha channel, it will make an attempt at correct alpha blending, so you can fade stuff out (a process called feathering or mask feathering), rather than having a sharp discontinuity from totally opaque to totally ignored. This should result in better output, because discontinuities are usually bad, especially in large-scale gradients like the blue sky, where discontinuities are very noticeable.

Because pv uses (hopefully) correct alpha blending, the results should be very good, and masked-out bits should be even less apparent than with other techniques. This is where I'd like feedback, just to see if my hopes to have implemented a higher-quality image blending method are justified. It should even be quite possible to do more interesting things with the alpha channel than just erasing unwanted bits - in pv, the quality masks for the B&A image splining algorithm are *weighted* with the alpha channel, so you might put whatever content you want into the alpha channel to make things appear 'less strongly' where the alpha channel is 'less opaque'. This should open up a whole new set of possibilities, but we can discuss this some other time.

So: Do you have such images (with pto) as example??that??shows??this functionality?

Sorry, not just right at hand, but I think you get my drift. If I find the time, maybe I can come up with an example, please bear with me.

Kay

--
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/4642fd46-1c02-6b76-b492-195e0addeb76%40yahoo.com.

Reply via email to