An investigation on Off-line Bank Document Images Localization and Recognition using WBDP, Texture and Perceptual Features

Off-line automatic reading of handwritten cheque is a computationally complex process however it plays an important role in the financial risk management. Computer vision and pattern learning provide a superior solution to this problem. Previously lot of efforts were taken for recognizing diverse pitches of cheques using template-matching of an identical outline. These schemes could fail when different outlines are encountered. In this paper, a feedback based localization and recognition is proposed. Outline problem can be resolved by cheque information tree (CIT), which generalizes the localizing method to extract active-region-of-entities. Feedback weight based density plot (WBDP) is performed to isolate text entities and reading integral pitches by ensuing exact localization. Recognition is based on texture features with neural classifiers. Especially legal amount is recognized by both texture and perceptual features. A post-processing phase is invoked to detect the incorrect readings by type-2 grammar using Turing machine. The proposed system was experimented to read cheque and demand draft documents acquired from different banks. The results obtained in the off-line experiment show that suggest approaches can easily be deployed in the regular practice without any major design amendments.