A Study of Augmented Reality Framework Based on Document Image Retrieval Techniques

• Main Authors: Tsuen-fu Liu, 劉荃富
• Other Authors: Jia-Hong Lee
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/65226674883476530486

碩士 === 國立高雄第一科技大學 === 資訊管理研究所 === 100 === In this study, we have built an animation news system using document image retrieval and augmented reality techniques. The goal of the system is to provide users with realistic audio-visual content when they are interested in reading some topics on newspapers. In our approach, the newspaper’s pictures were used as markers of augmented reality. The users can access these pictures to watch the videos of the news and to obtain more multimedia information. This effect is just as Harry Potter-style newspapers. Many image processing skills were applied in our system such as image segmentation, image recognition, tracking and augmented reality engine with audio-visual contents. In image segmentation process, Otsu thresholding and Run-Length were used to estimate the locations of four corner points of pictures. Then Hough transform was performed to find more correct locations for these four points. In image recognition process, image calibration was performed and Canny edge detection was used as features to identify different pictures on newspapers. Finally, four-point transformation was reused to augment video image frame on the image captured by a camera. In addition to make the system become faster, we used Full-Search tracking instead of image segmentation. Once the photographs have been identified, Full-Search process will track the four corner points of this picture to augment the corresponding video image frame on it. To evaluate the performance of the system, we collected 100 images from newspapers as a test set and the accuracy rate of the system can achieved 96 %.