http://swrc.ontoware.org/ontology#TechnicalReport
因子分解法による連続画像中の物体認識手法の改良
ja
NTTデータ通信（株）
マサチューセッツ工科大学
高野 秀也
アラン・ブリック
An improvement on the factorization method for computing three dimensional shape and motion from an object in a sequence of images is discussed. The original method requires correspondence between the feature points overall frames in the sequence. When feature points can only be detected in some frames (as in occlusion for example) a severe limitation is imposed to the original algorithm. We propose in this article an improvement on the factorization method that overcomes the above limlitation by adapting the original method on the partial frames in order to compute 3D shape and motion and integrating them into the final result.
An improvement on the factorization method for computing three dimensional shape and motion from an object in a sequence of images is discussed. The original method requires correspondence between the feature points overall frames in the sequence. When feature points can only be detected in some frames (as in occlusion for example), a severe limitation is imposed to the original algorithm. We propose in this article an improvement on the factorization method that overcomes the above limlitation by adapting the original method on the partial frames in order to compute 3D shape and motion and integrating them into the final result.
AA11131797
情報処理学会研究報告コンピュータビジョンとイメージメディア（CVIM）
1992
7(1991-CVIM-076)
41-47
1992-01-23