enblend mask optimization failures

Bug #880605 reported by Brent Townshend
10
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Enblend
New
Undecided
Unassigned
Hugin
Invalid
Undecided
Unassigned

Bug Description

It seems like the new seam line determination using graph cut has some problems during the optimization phase causing loops in the seam line that sometime result in omissions of parts of the image.

For example, see the attached seam line visualization.

Revision history for this message
Brent Townshend (bst) wrote :
Revision history for this message
Brent Townshend (bst) wrote :

I'm also uploading enblends verbose output. It looks like for this image pair, it thinks there are 5 seams to optimize rather than just 2 as there should be.

affects: hugin → enblend
Revision history for this message
Brent Townshend (bst) wrote :

I think this may be related to the --wrap=both option to enblend. One of the images in this particular blending had part of its extent wrapped around the edges of the image file. I think that's whats causing enblend to think there are more seams and coming up with the bad result.

By the way, the bad result persists with seam optimization turned off.

Revision history for this message
Brent Townshend (bst) wrote :

The failure occurs with or without the --wrap=both or --wrap=none. But it still appears related to the image overlaps crossing around the panorama boundary.

Revision history for this message
Mikolaj Leszczynski (rosomack) wrote :

You are correct about the loops. They are caused by the seam line optimizers, as confirmed by Christoph Spiel. The optimizers require fixes to cope with the new algorithm in some cases. As a temporary workaround, one can use a fine mask or the old primary seam generator, which is better compatible with the optimizers.

Seam line wrapping is not yet supported with the graph-cut algorithm, so it may or may not be the cause of this behaviour.

If possible, please attach the images for debugging.

tmodes (tmodes)
Changed in hugin:
status: New → Invalid
Revision history for this message
Brent Townshend (bst) wrote :

Why was this changed to 'Invalid'?

Revision history for this message
tmodes (tmodes) wrote :

@Brent
The status was changed to 'Invalid' only in the Hugin bug tracker (because this can't fixed in Hugin). The ticket is still open in enblend bug tracker.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.