One thing I miss in all the displays like masking, adding control points, etc: either a Photoshop-style drag tool for dragging a zoomed in image around in the viewing frame, or the GIMP style "Click on box in lower right corner of image window, get a small version of the image with a highlighted box showing the current window view, and drag the highlighted box around to move the image." (Sorry for the wordiness, I don't know what it's actually called.)
It sounds to me like your workflow too is a big work-around for a buggy masking process.
Am Mittwoch, 22. Oktober 2014 14:49:54 UTC+2 schrieb John Muccigrosso:It sounds to me like your workflow too is a big work-around for a buggy masking process.
It sound for me that the issue is behind the keyboards and stems from a wrong assumption, and not from a buggy implementation
The mask tab is not a fully fledged vector editing program with sub pixel accuracy. And this is not needed.
The seams are finally placed by enblend at the end. The masks are a help for enblend for placing the seam.
So there is no need to place the mask exactly. Often a very rough mask (with 4-10 points) is sufficient. Complex masks with many points are not needed. So the mask tab was written with this background. And it works for this use case.
See for instance the tutorial on Hugins website: The example for the include mask shows a nice and easy mask. Such masks are sufficient for the most cases. The example with the exclude mask I would consider already as border case, this should be the finest mask needed by Hugin/Enblend. Handling more complex mask is not the task of the mask tab. You can try it, it may work or not. But this is not the task of the mask tab and it is a misuse of the feature.
No one is asking for sub-pixel accuracy. We're trying to understand how the masking functionality works and noted that it behaves in unexpected ways and not as the manual (which is admittedly minimal) indicates.
Note however that the tutorial you like has a mask with over 30 points, and that it also includes the directions on how to manipulate masks that don't behave as described.
Am Mittwoch, 22. Oktober 2014 18:59:07 UTC+2 schrieb John Muccigrosso:No one is asking for sub-pixel accuracy. We're trying to understand how the masking functionality works and noted that it behaves in unexpected ways and not as the manual (which is admittedly minimal) indicates.
I can't reproduce the issues you mentioned. I can get only something similar as the mentioned issues if I use masks with very narrow points. And for this the mask tab is not indented.
Also the tutorial states this already:
cite: "Blend masks are not like your normal 'cut out' and 'paste in' masking. They are more like giving hints to the blender, so it isn't always necessary to carefully define a mask exactly on the boundary of an object. Often it sufficient to only roughly enclose the object to be included or excluded."
> I'd go so far as to say that it behaves unlike any similar function I've used in other software.
Then I have used other software than you.
Note however that the tutorial you like has a mask with over 30 points, and that it also includes the directions on how to manipulate masks that don't behave as described.
Where is this described? I don't see such text where issues like in your first post are mentioned.
PS: The selection tools use a little fuzzy distance, otherwise you would have to click exactly on the point or line, which would be nearly impossible for lines. Alternatively the selections points and lines would have to be drawn significant bigger and would hide parts of the image.
I'm not sure how narrow my points are. Too narrow, it seems. That may be the whole problem I'm seeing: distances are so small compared with what Hugin expects, that the fuzzy-distance logic is kicking in. In my case, my mask is about 100 pixels high and about half that in width.
My workflow is more based on suppressed-perfectionism combined with a
belief that a mask in Hugin specified "include/exclude this specific
area". Apparently that's not how Hugin uses masks; apparently it's more
of suggestion to the *blender* to include/exclude the area if possible.
My bad, thinking in terms of graphics software layers.
By excluding an area with a mask (an ""exclude"" mask), you are telling the blender to exclude that region from consideration both when placing the seam and when blending. The result is that you can remove a person or object from the final stitch. You should only exclude a part of one photo if some other photo shows the same part of the scene without that object, otherwise you will get a black empty area in your final panorama.
When you include an object with a mask (""include"" mask) the opposite happens, and that part of the scene is removed from all other photos. The blender has no choice but to use the region in the photo you selected.
That last bit about the blender having no choice seems a bit strong in light of Tmodes' explanation.
The tutorial on masking is more accurate, as noted already.