mirekslouf
commented
11 months ago - This issue NoiseReduction has its own branch ncorr and module idiff/ncorr.
- Previous results suggest that it is enough to process central region of each datafile = 160 x 160 pixels.
- it is better to suppose that central spot is approximately (but not exactly) in the geometrical center
- the neural network reducing the noise should take into account all datasets (SNR: perfect, good, average, and bad)
- later we should add an example some very bad dataset (SNR: poor)
Pavlina, reduction of noise on 2D-STEM micrographs from TimePix detector.