Normalise¶
The aim of normalise plugin is to normalise the output of the specim camara. This would give a dataset that can be compared with other datasets taken at different time, location or camara.
Theory¶
The normalisation plugin takes three samples, a grey reference, a black reference and a sample to normalise. To normalise the image a few steps have to be taken:
- Noise should be removed from the samples. This is done with the black sample, because it is taken with a closed shutter the images only contains noise.
- The black sample and sample are never taken at exact the same time. This means there can be a difference in (thermal)noise especially with large shuttertimes. This can be solved by subtracting a offset of the black sample so the noise levels are the same. This can be done in the settings. There is also a auto fill button, this only gives an indication what the offset can be. This number gives as a result that 98% of the top and bottom part of the images contains possive numbers.
- The data should be changed from raw intensities to amount of reflection (this is always a number between 0 and 1). This is by dividing the sample with the grey sample. The grey sample is a sample taken of a surface where the reflection is known, preferable flat across the whole spectrum.
- To make the data comparable with other samples that are normalised with different grey samples the grey sample must be made into a full reflection sample. This is done by dividing the grey sample by the data provided by the person that made the grey samples or by a number that is filled in by the user during the import of the grey sample.
- The environmental factors such as non homogeneous lighting must also be removed but that is done in step 3 by diving the sample with the grey sample.
The result is the following equation:
\[\frac{Sample - (BlackRef-Offset)}{(GreyRef - BlackRef)*DRT}\]
The normalised sample is then returned by the plugin into the project folder.
Depending on the type of image either a line-scan or area-scan algorithm is used, both algorithms use multiprocessing.
Using the plugin¶
- Load the plugin by selecting the grey, black and raw images (.TippHSI).
- Fill in the filename and when desired the offset (this can also be negative).
- Run the algorithm and the output is saved and can be visualised with the normal visualisation plugin.
Remarks¶
Somethings to take into count when normalising:
- The black and white images should always be made under the same conditions and at the same time as the samples. Also they should be made directly after each other.
- A lot of negative datapoints can lead to serieus errors in futher analysis especially when doing something with the difference in intensities, because negative numbers will be stored with modulo so they become very large.
- The offset only corrects for the sample, beacuse the grey should be made directly in front or after the black, with the result that there will be no offset between de black and grey.
- To check if the offset is correct select some points in the normalised image and check if the lower and higher bandwidths do not have both or a large positive or a large negative sloop.