Preprocess ID Card image for Tesseract OCR

I have a script that :

recognize ID card
crop image
recognize ID fields
extract text from fields

but has some problems

1. in some conditions fails extracting text on fields
2. single character field (sex M or F) is not being extracted

I’m only doing gray image preprocessing , I need a more complex filtering to improve the extraction

I’m using OpenCV + pytesseract