Normalization¶

Normalization using ROI (optional)¶

If you want to specify a region of your sample to match with the OB

Let’s use the following region

• x0 = 10
• y0 = 10
• x1 = 50
• y1 = 50
>>> my_norm_roi = ROI(x0=10, y0=10, x1=50, y1=50)


then the normalization can be run

>>> o_norm.normalization(roi=my_norm_roi)


Normalization without ROI¶

If you don’t want any normalization ROI, simply run the normalization

>>> o_norm.normalization()


How to get the normalized data

Each of the data set in the sample and ob will then be normalized. If a norm_roi has been provided, the sample arrays will be divided by the average of the region defined. Same thing for the ob. Those normalized array can be retrieved this way

>>> sample_normalized_array = o_norm.data['sample']['data']
>>> ob_normalized_array = o_gretting.data['ob']['data']


Forcing normalization by mean OB¶

By default, if the number of sample and Ob is the same, each sample is normalized by the equivalent index ob. But it’s possible to force the normalization by the mean OB

>>> o_norm.normalization(force_mean_ob=True)