peempy

A suite for processing PEEM-XMCD data from I06 Diamond Light Source.

PEEM-XMCD work flow

XMCD signals are from the difference of images taken under two different circular polarisation of X-ray. There are two types of data - XMCD4 and XMCD2. The former captures 4N images of resonance+, off-resonance+, resonance-, off-resonance-. Usually N=10. Images of each set are averaged. The polarisation dependent signal is taken from the difference between on and off resonance.

The difficulty is that there is drift over time but averaging the signal is required to reduce noise. The images need to be drift-corrected to minimize mismatch between them before taking the average. In addition, on and off resonance images need to be drift corrected separately.

What this does

Drift correction for each capture

A more robust drift correction scheme is implemented, taking the advantage that the drift is aggregation process. In the original Igor implementation, an ROI containing a feature needs to be selected and the feature has to stay in the ROI in the entire series. Here, we allow a floating ROI for each frame. The ROI of the next frame is shifted based on the calculated absolute displacement of the current frame. This effectively makes the ROI "lock on" to feature in the image series. Hence, the initial ROI only need to be as large as the feature, given that the relative drift between the frames does not move the feature completely out of the ROI.

The routine in skimage.feature.register_translation is used. It is much faster than the standard (super-sampled) image convolution scheme. The reference image is improved on-the-fly by adding weight averages of the corrected frames for noisy data.

Batch processing

A script for batch processing has been written. Using python's multiprocessing module, parallel batch processing can be performed to use all computational power. Captures can be previewed before processing for selecting a sensible drift area. For memory efficiency, it is desirable to use the inplace drift processing model to avoid creating addition copies of the data.

XMCD signal calculation

XMCD signal image can be generated using peempy.xmcd

XMCD vector map construction

XMCD vector map can be generated from the signal of more than one of the angles. Due to the distortion from the instrument, frames need to be aligned before fitting. This package allows image alignment between an arbitrary number of frames based on any number of control points for alignment. Three-dimensional vector map can be constructed if XMCD images are captured with more than three angles.

TODOs

• Write quick-start documentations