What image preprocessing can I use to make my image google vision OCR work better?

Question

I am using google vision document_text_detection on this image:

I am trying to read the digits in the middle. Now the google vision API works wonderfully well