Right now, DataLayer does image-specific processing of LevelDB data.
It should be separated into LevelDBDataSourceLayer, and ImageProcessingLayer.
That way, other data sources can more easily be added: HDF5, image directories, CSV, etc.