Build a OCR Linux command line software

Goals To create a linux command line interface software that receives as arguments a PNG/JPG image file and a regular expression and outputs the recognized characters validated by the regular expressions. Summary We have a software that uses feature matching to identify different types of documents and that find the specific text regions that we want to read. This software crops these text regions and generate the PNG/JPG files of the text regions that should be recognized. We tried to use tesseract but the results were not good enough. We believe to improve the results what should be done is: - Pre-process the images to improve OCR results (deskew, 1-bit, unsharp masking); - Use the regular expression on the input to improve the OCR process; Objectives Develop a command line tool that receives an PNG/JPG image file and a regular expression and outputs the text on the image validated by the regular expression; Deliverables Command line software to run on Amazon Linux (AWS Lambda) or Ubuntu; Basic architecture that should be followed: Arguments: - image file path; - regular expression; Output: - text result; Details 100 images will be provided for testing purposes;