ks-affine-aligner

An OpenCV-based Python tool for alignment of (presumably clear-skull) images.

How it works

The align_image() method takes a set of images, and works in a following way:

1. The center of mass (CM) is computed for each image, and the template (reference) image is picked up based on the A-P position of the CM (the actual percentile of the reference image among all the input image may be specified using the use_percentile option).
2. Gaussian filter-based background subtraction is performed on each image. One can specify the width of the filter using the background_dia option (in pixels).
3. The ORB feature points are extracted from each image, and alignment is performed between the template / reference image (one may be able to tweak this step via changing the feature_size and threshold_factor parameters).
4. A 2-d affine transformation matrix is estimated for each image, based on the alignment of the feature points.

Other notes

• The Coordinates class represents the IJ coordinate space of the image. Pre-computing one may speed up the procedures to a certain extent.
• The routines in the compute submodule may be useful for other image-processing purposes.