When you are using FeatureSpacePlot you can use options like FeatureExtractor and Method to customize the type of features being extracted from the data. However FeaatureSpacePlot does not immediately make available the extracted features.
To look at the extracted features, you could use FeatureExtract (https://reference.wolfram.com/language/ref/FeatureExtract.html) and DimesnionReduce (https://reference.wolfram.com/language/ref/DimensionReduce.html).
What FeatureSpacePlot provides is a quick and easy alternative to using FeatureExtract on your data followed by DimensionReduce (to reduce number of dimensions to 2) and then use ListPlot.
You can try the following code:
imgs = Rasterize /@ Alphabet[];
FeatureSpacePlot[imgs]
ListPlot[Thread[DimensionReduce[FeatureExtract[imgs], 2] -> imgs]]
The two visualizations may be different because we are not specifying any options in the code above and the methods picked during the two sets of operations may be different with defaults settings. You could possibly get similar results by fine tuning options (Method, FeatureExtractor, etc.) for both sets of functions.
