It is possible to explicitly train the background in tesseract 4

[Sebastian Correa]

I am developing an OCR system and I am interested in being able to train the model in tesseract4 by explicitly describing the background (image or color) in addition to the boxes-image-charcater files. Is it possible to do this? or is implicit in the LSTM-based model training with the boxes?

Thank you
Sebastian