Preprocessing and presorting of envelope images for automatic sorting using OCR

摘要

This paper reports a review and feasibility study undertaken to investigate methods of preprocessing envelope images to (i) extract the address block from the image in the presence of other data, and (ii) presort the addresses into sub-classes suitable for recognition by an OCR system with separate recognition channels for machine and handwritten address classes. Results are reported from an analysis of over 1500 actual address images of different print and writing styles obtained using existing OCR sorting equipment. These results show a correct address block location rate of around 98%, and that presorting of addresses into machine printed or hand printed/written was around 86% correct. Sorting in terms of OCR machine readable or not OCR machine readable (as defined by existing OCR equipment) was around 70% correct.