Control points are not generated reliably
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Hugin |
Invalid
|
Undecided
|
Unassigned | ||
Bug Description
I am trying to generate control points for a stack of images (which I intend to process into a time-lapse sequence). All images were taken from the same position (save for some camera shake) at intervals between 1 and 4 seconds, thus there are plenty of control point candidates.
I import them (using the Advanced interface) and select a batch of 120 consecutive images (as I get an error about to many arguments being supplied when the number of images approaches 127), then I generate control points.
When I examine the control points, I notice that sporadically there are pairs of consecutive images that do not have any control points between them. I have to go back to the first tab, select those two images separately, then regenerate the control points again (with the same settings as before). After that, the control points are there.
This has happened on every sequence I have processed so far. In 120 images, I usually have around a dozen image pairs without control points. Since explicitly regenerating control points works, the images are similar enough for Hugin to recognize similar features in them. It looks like the script which generates the control points occasionally skips some of the images it iterates over.
I use the "Align image stack" script to create control points. The Hugin version is 2015.0.
A full description of the procedure I follow is at http://
| Michael von Glasow (michael-vonglasow) wrote | #1 |
| Michael von Glasow (michael-vonglasow) wrote | #2 |
Looking at another project with the same issues, I see that image #0 shares no control points with #1, but it does share some with #6 and #11. It appears that the order of images gets messed up as they are passed to, or processed by, the feature matching script.