I-028 Sparse Decomposition of EPI by Using Greedy Pursuit Algorithm

概要

論文の詳細を見る
EPI (Epipolar Plane Image) is the basic element in representation of 3D image and FTV (Free-view point TV). In order to generate the FTV, a huge amount of data has to be acquired, which has been posing a great challenge for the application of FTV. Numerous methods have been conducted to handle the capturing issue until the concept of compressed sensing was proposed, which has triggered a great revolution in signal acquisition field. Fortunately, the compressed sensing can also be used for capture of Ray space to generate FTV. In addition, for reducing the number of measurements in compressed sensing, it is necessary to analyze the sparsity of signal first, and the special properties of EPI give us a platform to do the special process and analysis to this category of image. In this paper, we focus on the sparse decomposition of EPI, and try to represent EPI by the combination of few atoms in the overcomplete dictionary by greedy pursuit algorithm. The experimental result gives the comparison of the sparse decompositions by using different dictionaries.
FIT(電子情報通信学会・情報処理学会)運営委員会の論文
2012-09-04