Scanned documents can have pages with wrong alignment, low contrast, low brightness and rotated upside down. This can create challenges while processing of documents using OCR, ICR, Text extraction, image-based ML/AI modelling, etc. This solution incorporates statistical models which identify angle of tilt based on textual orientation and position of text relative to page boundaries and corrects the alignment / tilt of the pages. It identifies the contrast between background and text in the pages of the scanned document and adjust the contrast of the low contrast pages. It also incorporates deep learning models which identify if a page is upside down. The models are trained on a large dataset of thousands of pages. This enables OCR/ICR engines to achieve higher accuracy and improves the subsequent text extraction pipelines.