Ask Your Question

stitch transform reuse

asked 2020-12-01 05:23:14 -0500

nico50 gravatar image

Hello, i'm working with drone data, and special camera's, for agricultural data. Essentially I have a 5 channel camera, so I have more data than just RGB data (3 channels). So I ave something like "RGBTH"

I would like to be able to stitch images (for simple RGB that's not a problem I manage that with python). Though I would like to put in the other channel data at the matching positions.

Therefore I wonder is it possible to somehow export the RGB stitching transforms, and then re-apply it to the other camera layers, I would then be able to output stitched RGB (as normal) create another exactly overlapping output of the same image with TH data layer. (just putting T in R, H in G, and for B I'll use a grayscale of the normal RGB. The TH data is just stored as another tiff file set like the RGB data set, it has the same pixel resolution.

The problem I don't get the transform numbers out of it and don't know if it's possible to reuse them to stitch I'm using python btw.

edit retag flag offensive close merge delete


just curious, what's "TH data" ?

can you show us example RGB and THgray images ?

berak gravatar imageberak ( 2020-12-01 05:39:54 -0500 )edit

it tells something about specific plant reflection to certain wave lengths, its still very experimental though. The camera is in development as well, so I can though not show what it can do at this time sorry.

nico50 gravatar imagenico50 ( 2020-12-01 09:37:36 -0500 )edit
supra56 gravatar imagesupra56 ( 2020-12-01 13:38:03 -0500 )edit

that's for 5 images/videos, not 5-channel data

crackwitz gravatar imagecrackwitz ( 2020-12-01 21:10:16 -0500 )edit

I have a 5 channel camera

supra56 gravatar imagesupra56 ( 2020-12-02 08:12:31 -0500 )edit

precisely, the link doesn't apply to the situation

crackwitz gravatar imagecrackwitz ( 2020-12-02 14:12:33 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2020-12-01 11:22:07 -0500

crackwitz gravatar image

updated 2020-12-01 11:25:19 -0500

I took a quick peek at the stitcher API

it recommends that you familiarize yourself with its pipeline

it appears that you can run estimateTransform() separately from composePanorama(), with different images. give it a try.

you would split your 5-channel image into its planes (numpy slice, img[:,:,0] for first channel plane; if the api complains, try making a copy of the plane to make it contiguous in memory), run them through the composition method, then reassemble them (np.dstack)

edit flag offensive delete link more


it seams a deeper level of code, the basic stitching examples don't seam to use estimateTransform() and composePanorama() ??. Currently, I have it split as two tiff file sets. If somehow estimateTransform() on RGB can be made to work then the same transform data can be used on the TH image set. Do you know of some example code with estimatetrnsform maybe printing output or so?

nico50 gravatar imagenico50 ( 2020-12-01 14:33:16 -0500 )edit

OpenCV comes with simple and detailed stitching sample source and accompanying, very brief, "tutorial" article

crackwitz gravatar imagecrackwitz ( 2020-12-01 21:09:03 -0500 )edit

Question Tools

1 follower


Asked: 2020-12-01 05:17:29 -0500

Seen: 45 times

Last updated: Dec 01 '20