fdarvas
commented
2 years ago For shots at close distances, a band-to-band alignment of different channels - such as e.g. red to green - requires the solution of a dense stereo matching problem.
A method for this is described here:
High-density stereo image matching using intrinsic curves December 2018ISPRS Journal of Photogrammetry and Remote Sensing 146:373-388 DOI: 10.1016/j.isprsjprs.2018.10.005
Ideally this would be done between two channels that are most similar in the spectral response (such as green and blue) and are the most separated.
If a solution for the dense matching is found, i.e. basically a full depth map for each pixel in e.g. the green channel is available, a mapping to the other bands can be found via construction of the trifocal tensor. At this point, only the camera rig-relatives and translations are required, which can be found in the image metadata.
This is described in detail here:
Shahbazi, M., and C. Cortes. "Seamless Co-Registration of Images from Multi-Sensor Multispectral Cameras." The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences 42 (2019): 315-322. https://pdfs.semanticscholar.org/dff5/e50ebed805f18ca03ec9e95fdb5a4c507801.pdf
This is a rather complex problem and thus we will not support it in this toolbox.
wanghanruibaba
I am using the MicaSense ® RedEdge camera to take close-up pictures. Due to the various imaging systems in each channel, a phase difference is generated in the captured 5-channel images. How to align and register the 5-channel images, looking forward to your reply ,thanks