Kameda, Research, Documents and Info

Conclusion

We proposed a new model-matching method to estimate the pose of an articulated object from only one silhouette image. We have separated the model and the algorithm clearly. The matching method refers the contour relationship. We have showed its availability throughout the experiment for computer generated images. We also applied it for a real-life human by introducing the gnawed image and presented several experimental results.

If the method can control the order of the parts to be processed with some reasonable criteria, it will not only reduce the computational amount but also remove the uncovered silhouette region problem. This issue will be studied in the future.