Control points improvement proposal

86 views
Skip to first unread message

E Kow

unread,
Jan 5, 2024, 7:14:49 AM1/5/24
to hugin and other free panoramic software
Hi,

I am normally stitching around 500 (25horizontal x 20vertical) microscope images into one overview image. Hugin is doing a great job, but creating the controlpoints is not optimal. 
My images are named "image_1_1", "image_1_2" etc, all the way up to "image_25_20".
If I run Hugin CPFind with default settings, it will compare all 500 images to each other, which takes several hours. What I normally do is that I add the argument "--linearmatch" to CPFind, which then finds all the necessary control points for images that are horizontally next to each other. However, I also need controlpoints for vertical images, and I have not been able to find a good workflow for this. What I currently do is to manually select all the images that end with "_1" and run CPFind, manually select "_2" and run CPFind etc. It takes a lot of time to select images on every 25th row in the photos list, and repeat this at least for every second vertical column to have a good result.
I guess what I am looking for is something like --linearmatch, but it should compare every 25th image. 

Bruno Postle

unread,
Jan 5, 2024, 7:54:39 AM1/5/24
to hugin and other free panoramic software
I can't check the exact usage from here, but look for the 'multirow' option.

This does a linear match (matching consecutive images), then matches the ends of each resulting disconnected chain together, then finally does a 'prearranged' match (with the hope that the previous steps have put everything in an approximate position).

T. Modes

unread,
Jan 5, 2024, 10:13:24 AM1/5/24
to hugin and other free panoramic software
bruno...@gmail.com schrieb am Freitag, 5. Januar 2024 um 13:54:39 UTC+1:
I can't check the exact usage from here, but look for the 'multirow' option.

This does a linear match (matching consecutive images), then matches the ends of each resulting disconnected chain together, then finally does a 'prearranged' match (with the hope that the previous steps have put everything in an approximate position).

Alternatively you can assign rough positions to the images (via template or pto_var) and then call cpfind with --prealigned option to only match now overlapping images.

 

E Kow

unread,
Jan 11, 2024, 6:17:08 AM1/11/24
to hugin and other free panoramic software
Thank you guys. 
First running CPFind --linearmatch and then running CPFind (Prealigned) is a much more efficient procedure than what I have done before. 
Reply all
Reply to author
Forward
0 new messages