In this SBIR we propose to demonstrate the technical feasibility of Mobile OCR, a portable software system which makes use of existing personal devices to provide access to textual materials for the elderly or the visually impaired. The system will help these low vision individuals with basic daily activities, such as shopping, preparing meals, taking medication, and reading traffic signs. It will step beyond our proposed MobileEyes vision enhancement system to apply cutting edge recognition technology for mobile devices. The system will use common camera phone hardware to capture and enhance textual information, perform Optical Character Recognition (OCR) and provide audio or visual feedback. Our research will focus on implementing and integrating new vision enhancement and analysis techniques on limited resource mobile devices. Specifically, we will develop algorithms for detection and rectification of text on planes and generalized cylinders subject to perspective distortions, implement more robust and efficient algorithms and systems for stabilization and enhancement of text blocks, provide mobile OCR on complex textured backgrounds, and implement these techniques on small devices across a variety of platforms. The recognized text will be presented through Text-to-Speech (TTS), or displayed on the device with enhanced quality which can be easily read by low vision users. Phase I will focus on demonstrating the technical feasibility of our approach, and will incorporate a performance measurement methodology to quantitatively evaluate progress and evaluate our system against other approaches. In comparison to existing vision enhancement devices, such as magnifying glasses, telescopes, and text reading devices such as scanner-based OCR, our solution has several advantages: 1) it makes use of a single, portable device (camera cell phone) that is commonly available and typically already carried for its telecommunications capabilities; 2) it can be used selectively by users so they will not be overwhelmed by irrelevant information; and 3) it can be integrated directly with other applications for specialized tasks. Our research results will impact the millions of low-vision individuals and the blind, as well as vision and computer vision researchers. Our team is uniquely qualified to explore the feasibility of extending visual applications to these devices, and provide a platform for integrating future vision algorithms
