Global Image Registration using Random Projection

概要

論文の詳細を見る
Using the efficient random projection, we develop an efficient algorithm that establishes global image registration. Image registration, which is the first process in image data alignment and applied in remote sensing, image processing, medical image analysis, robot vision, car mounted vision system, and industrial image inspection. The current image registration techniques are classified into local registration and global registration. Our aim in this paper is to speed up global registration using random projection. Random projection refers to a simple technique of projecting a set of points from a high-dimensional space to a randomly chosen low- dimensional subspace and can preserve the geometric properties in the higher-dimensional space. We develop a global registration using random projection We use spectrum spreading and circular convolution to reduce computational cost of random projection.
2011-05-12