作者: M. Shizawa , K. Mase
关键词: Minification 、 Optical flow 、 Artificial intelligence 、 Flow (mathematics) 、 Process (computing) 、 Dimension (vector space) 、 Mathematics 、 Convolution 、 Pattern recognition (psychology) 、 Algorithm 、 Computer vision 、 Energy (signal processing)
摘要: The authors propose a simultaneous closed-form estimation method for multiple optical flow from image sequences in which each point has motions. This only requires convolution space-time filtering and low-dimensional eigensystem analysis as an optimization process. mixture model of energy integral minimization fitting method. It is shown that symmetry between component flows the can reduce dimension make unimodal stable. Successful experiments on double-flow random texture patterns natural scene images are reported. >