Start new module for feature extraction

[MaximLippeveld]

Features should be extracted from masked and unmasked features.
The feature extraction modules will take a Dask bag as input (output from segmentation) and a Dask DataFrame as output.

• Compose list of features to be implemented (what libraries to use)
• Implement extraction module per feature type (eg intensity, shape, texture ...)
• Embed feature extraction in main pipeline

[sanderthierens]

For the shape features add multiple non connected masks handling

[sanderthierens]

Possible feature Histogram spread dor image contrast metric

code:https://stackoverflow.com/questions/63437029/implementing-histogram-spread-for-image-contrast-metrics
paper:https://www.researchgate.net/publication/216187588_Performance_Metrics_for_Image_Contrast

[MaximLippeveld]

Dask recommends storing data in the Apache Parquet format. This format also gives good interoperability with other languages and frameworks. https://docs.dask.org/en/latest/dataframe-best-practices.html#store-data-in-apache-parquet-format