Ask Your Question

Stitch images by Translation

asked 2016-08-10 04:11:06 -0500

Ael gravatar image

I am currently using OpenCV stitcher class to stitch two images acquired by a camera translated parallel to a wall. The distance between the wall and the camera is around 75 cm with a limited field of view (around 70 degrees). I am using the Stitch() function to stitch such images and it works well for most of the cases. The camera is moved manually thus there may be also some rotations in the process. I know that these are taken into consideration by OpenCV in bundle Adjustment but what about the Translation ? Is the stitching working because the scene is planar therefore no parallax is involved? Has anyone worked with OpenCV Stitcher to stitch images from a translating camera ?

edit retag flag offensive close merge delete


You can read this post and this reference

LBerger gravatar imageLBerger ( 2016-08-14 02:54:25 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2016-08-10 06:57:55 -0500

emiswelt gravatar image

updated 2016-08-10 06:58:17 -0500

OpenCV has no support for translation in stitching, since that's highly use-case specific.

You'll have to implement your own Stitcher that does alignment, warping, seam finding and blending. Have a look at the detailed stitching example. The OpenCV example does a lot of tuning (linke calculating seam finding etc. on smaller images) which you can throw out.

If you really need to use bundle adjustment for translational alignment, you will have to implement your own version of BA. Usually you will be able to use simpler methods for most use-cases. You will also need to implement your own warper to project the images to your panorama's target space (probably a plane).

edit flag offensive delete link more


I am in fact using the stitching_detailed example. The scene is planar (a wall) so I assumed it would still work out as it uses homography which holds for such a scene. It is in fact functioning properly for a considerable number of image pairs however there are some others for which it is not working. I am realising that this happens when the matching confidence is very low, thus assuming it is actually a problem with the features and their matching. Some images do in fact seem to very undistinctive features when viewed visually....

Ael gravatar imageAel ( 2016-08-14 02:28:37 -0500 )edit

Question Tools

1 follower


Asked: 2016-08-10 04:11:06 -0500

Seen: 1,371 times

Last updated: Aug 14 '16