How would you improve the receipt recognition for this specific receipt (images are included)?

Currently, we are doing some pre-processings and use tesseract to extract text from this particular receipt.