I have extremely large reference and target satellite images (4 bands each) which I want to tie using SIFT. I am exploring skimage for this purpose.

My strategy is to use a ‘chunk-based’ approach to handle the image sizes. Currently, for each chunk I run the matching process and saving the results per chuk.

However, I want to test a different approach. I want to run the match process for all data, for whole image, in one go. How could I combine SIFT points and descriptors for all chunks, for both the reference and target images and run the match at the end for whole image and warp the target image, band by band.

Thanks in advance.

Maz

I used python and skimage library for SIFT matching on two extremely large images. The process is very slow. I want to optimize.