Absorption images are numbered by sequence time and taken in sets of 3:
end letter A: absorption image with atoms present
end letter B: reference image of illuminating light field, without atoms
end letter C: reference image of background illumination, without laser light

An image sequence is analyzed using principal component analysis, i.e. by reconstructing the most suitable background image from the set of B-images, thus accounting for typical fluctuations (moving interference fringes due to vibrations). The most likely background matching an A image is found by projecting the image onto the most dominant principal components of the B-set.
