Abstract:
This paper integrates a time-of-flight (ToF) sensor to a high-resolution charge-coupled device camera in a new and fully automated book reader design for individuals with visual impairment and blindness. This cost-effective and portable book reader design integrates, in a unique way, a 3-D map extracted by a low-resolution ToF sensor and a high-resolution image as means to correct the page curvature of bound books. This is achieved by a seamless registration process of two imaging modalities, so that the low-resolution (160 × 120 pixels) depth map accurately reflects the entire book spread as captured by the high-resolution camera (3072 × 2304 pixels). As part of the automated registration process, the book reader will give the user a warning message if the book is misplaced within its field of view. The reading accuracy is further improved by a mathematical framework formulating the flattening and extension operations, which together optimally dewarp the book spread images. This correction process is resilient to potential misplacement of the book within the book reader platform. The testing phase led to a reading accuracy of over 95%, involving an unprecedented acquisition and processing of 200 pages. A database consisting of all the 200 high-resolution book spread images together with their respective extracted depth maps are made available to the research community for further developments.