High Speed Book Flipping and Scanning

We are increasingly accustomed to using e-books and each book is now released in a digital version too. Progressive digitization of the world around us causes the demand for devices to enable the digitization of paper books, official documents, and notes. However, the conventional technology can not meet the demands for ease-of-use and high-speed book digitization. Ishikawa Oku Laboratory at the University of Tokyo works on a solution.

There is a growing demand for camera-based document analysis and recognition and there emerges a solution to the speed of book scanning named BFS-Auto. This device can achieve high-speed and high-definition of a book digitization at over 4 pages/sec. This performance is realized by three key points:

1. high-speed fully-automated page flipping,
2. real-time 3D recognition of the flipped pages,
3. high-accuracy restoration to a flat document image.

The mechanism can not cause obstruction for the scanner while maintaining full speed. The developed system can flip and scan the book at over 250 pages/min without modifying the book by cutting. The real-time 3D recognition system continuously observes 3D deformation of each flipped page at 500 times per second, and recognizes the best moment for book image digitization at the highest quality using the newly constructed original algorithm in real-time. At this recognized moment, the high-definition camera captures the document so that both high-speed and high-definition performance is realized. Finally, the 3D restoration system restores a captured image which is distorted because of page curling to a flat original document image by using the captured image and the obtained 3D deformation.

The technical details are not available because as Yoshihiro Watanabe, the leader of the project, said,

the detailed research paper has not been published and this work is collaborative research with the company so that we need to keep some details in secret.

See the video below, that shows this forward-looking and somewhat revolutionary machine in action. Optical character recognition (OCR) software will be an excellent complement for the system to convert scanned images of handwritten, typewritten or printed texts into machine-encoded texts.