Panoramic Scanned Page Using Mobile Phone Camera

Today, almost all mobile phones introduced in market are equipped with built in cameras and in this work, we have proposed a new application by using this camera. A scanned image of a large document is achieved by com-pletely traversing this camera on it. All successive frames are joined seamlessly to generate a panoramic image of the page. It is a difficult task due to, lateral movements of cam-era along with hand shaking, change in illumination due to hand shadow etc. Matter is further complexed by the fact that text alphabets does not possess distinct features which is a key requirement for panoramic image generation. A framework is designed for coarse to refine alignment of im-ages based on the background modeling approach. Fore-ground segmentation and correspondence estimation are expressed as a unified labeling problem, and are solved effi-ciently via tree dynamic programming (TDP). Experiments proved our algorithm to be robust in performance.