Great scanning and OCR for mobile devices

The aim of this project is to improve the scanning and optical character recognition on mobile devices. Currently the cameras of many mobile devices have relatively noisy output whenever lighting conditions are less than optimal. Additionally, it's almost impossible to achieve scans that are distortion free as mobile devices don't have a surface to which the document under scan could be pressed to reliably. These two problems lead to difficulties in performing optical character recognition over acquired images as most recognition algorithms require an input that is noise and distortion free. The solution that will be developed by this project will solve both of these problems by acquiring multiple scan images from different angles. Same objects can then be matched across the source images providing two benefits: the noise can be cancelled out and 3D shape of the document under scan can be derived. Such information can then be used to unfold the document to 2D space and provide a noise and distortion-free image to optical character recognition algorithms. The solution will be implemented taking into account the performance limitations of mobile devices and a major optimization effort will be spent to achieve an acceptable latency of the complex image processing algorithms.