時空間相関法によるオプティカル・フロー速度計測法の改良

概要

論文の詳細を見る
To date, several methods determining optical flow have been proposed for stereo vision (passive ranging of depth and shape perception). ”Gradient method” and ”Matching method” are well known and representatives among them. But, there are many restrictions and problems to apply these methods to actual scene analysis. Recently, two of our authous proposed a new method for the determination of optical flow with pixel-based spatio-temporal correlation analysis of sequential image. One can determine optical flow velocity with fine accuracy and high density by the correlation method. However, there still exists some problems such as it needs enormous time for calculation and it is difficult to analyze a passage of plane wave. Both of these are key problems for real-time stereo vision analysis. In this paper, we develop our correlation method and propose new algorithms to conquer these problems. Usefulness of the new algorithms is confirmed with artificial images and an actual dynamic scene (chemical wave propagations).