Augmented Reality: Synthetic Poster Insertion - トラッキングの状態が目で見て分かる#ARデモ
ARの画像認識分では、特徴点が...という会話がよく行われる。この特徴点をリアルタイムに解析し、そこには存在しない絵を表示しているデモストレーション。特徴点の解析状態が良く分かる。
This video showcases the application of our latest algorithm (published in CVPR 2012) for synthetic poster insertion into a video stream. Our algorithm detects multiple homographies (roughly, planar structures) given a set of point matches across two frames in the video, which allows us to establish the proper geometry for photo-realistic insertion of synthetic posters.
A major advantage of our approach is the ability of automatically decide the appropriate number of available planar structures, and the high efficiency of the algorithm (less than 1s per frame).
Our algorithm is described here:
T. Pham, T.-J. Chin, J. Yu, and D. Suter
The Random Cluster Model for Robust Geometric Fitting
In Proceedings of Computer Vision and Pattern Recognition (CVPR), 2012, Providence, Rhode Island, USA.
Download code and paper from http://cs.adelaide.edu.au/~tjchin.