Global RMS in Autopano

[Jeremy Payne]

Can anyone explain what the Global RMS score means/is in AutoPan Pro?

I often do a "test" of a big pano by using down-sized files first.  I just did that with a 15 shot bracketed Pano ... and with the small files (800 pixels on the long side) I get a Global RMS of 1.20.

When I send the same 15 images to AutoPano Pro at full resolution, I get a Global RMS of 2.28.

Why is that?  Shouldn't the score be the same?  Is there more room for measuring error with more pixels?

[AndrewKulin]

It has to do with the quality of the stitch/control point errors - the lower the number the better.  I think that you have pretty good numbers all around, why they change I cannot say.

You can also download the AutoPano user manual from there site.  In around pages 19-22 (+/-) their is some mention of RMS and Global RMS.  What jumped out at me was a figure showing a pop-up box from the software where you could apply stitch settings for the program to follow when stitching.  One of the options was a slider bar to "keep only control points below this error (RMS).  The range was from 2.0 to 20.

Andrew

[Panopeeper]

Can anyone explain what the Global RMS score means/is in AutoPan Pro?

This is not an Autopano speciality, the other stitchers may name it differently.

For judging the quality of the alignments of the image pairs, the distance between the selected matching points (control points) in the warped images is measured. Zero distance means, that the stitching is perfect. Parallax error, unrecognized or incorrectly specified lens geometric distortion, incorrect control point placement and error in the stitching process cause non-zero distances, i.e. misalignments. The RMS (root mean square) is the square root of the average of the squares of distances.

The goal of the "stitching" (alignment of frame pairs) is to minimize this value; the result is, that even perfectly placed and perfectly matching control points will be pulled apart in exchange for pulling very far control points together.

Placement of the control points is a critical step; I have been placing the control points manually for years.