Multi-modal text recognition and encryption in scanned document images.

Many military and business documents such as memorandums, invoices, medical records and bills among others are transmitted over the network in the form of images. These scanned document images may contain some private information that must not be accessed by third party or unauthorized users. Most of the encryption techniques encrypt the complete image but only parts of it require protection, which increases the computational and overhead communication cost. In this paper, we introduce a novel technique for selective multi-modal text recognition and encryption in scanned document images to overcome the issue of security and efficiency. Maximally stable extremal regions (MSER) are used in combination with two filtering techniques to automatically detect regions of interest (ROI) that contain text. Proposed technique combines optical character recognition (OCR) and natural language processing (NLP) in text recognition thus taking advantage of different modalities of data. Multi-modal recognized text is then encrypted using advanced encryption standard in cipher-block chaining mode (AES-CBC) and hybrid chaotic map. Simulation results illustrate the reliability and efficiency of proposed encryption scheme based on ROI. Moreover, security analysis exhibits the robustness of proposed encryption technique against linear and differential cryptographic attacks. [ABSTRACT FROM AUTHOR]