Document Image Rectification Using Single-View or Two-View Camera Input
Digital cameras have several advantages, for instance, portability and fast response, over flatbed scanners. Therefore, there have been a number of attempts to replace flatbed scanners with digital cameras. Unfortunately, camera captured images often suffer from perspective distortions due to oblique shot angle, geometric distortions caused by curved book surfaces, specular reflections, and unevenness of brightness due to uncontrolled illumination and vignetting. Hence their visual quality is usually inferior to flatbed scanned images, and the optical character recognition (OCR) rate is also low. Camera captured document images thus need to be enhanced to alleviate these problems and to widen the area of valuable text processing tools (for example, OCR and text-to-speech (TTS) for the visually impaired, automatic translation of books, and easy digitization of printed material) for the camera captured inputs. This chapter focuses on removing perspective and geometric distortions in captured document images, operations that are referred to as document dewarping or document rectification.