Panorama Mosaic Optimization for Mobile Camera Systems

In order to enable the real-time image mosaic in mobile devices that have very limited computing power and memory, we propose an efficient image mosaic algorithm with integer arithmetic. The proposed algorithm is focused not only on the computational efficiency, but also on the performance improvement of image mosaic. For the efficient projection without transform estimation, we obtain successive images through the semitransparent view finder which shows the fixed amount of right part of previously captured image. We project each images onto cylindrical surface by the proposed projection equation and integer arithmetic. We align the projected images by hierarchical hexagon search and a matching measure that is robust to illumination changes caused by different exposures and vignette distortion. Finally, the images are naturally blended using color compensation and dynamic programming based stitching. The proposed mosaic algorithm is embedded and tested in mobile phone systems. According to various tests and experiments, the proposed algorithm shows good panorama composition in real-time compared with the other PC-based methods.