科研之友
微信
新浪微博
Facebook
分享链接
摘要
A time sequence of high-quality images currently produced by high-resolution observations either from the ground or in space may be utilized to determine the transverse flow field on the plane of the sky with the help of optical flow techniques. We have examined the performance of three different methods-a well-known technique called local correlation tracking (LCT), a recently developed technique called the differential affine velocity estimator (DAVE), and a new technique called the nonlinear affine velocity estimator (NAVE)-using three kinds of image data: mapping-based synthetic images, a set of MHD simulation data, and real images (magnetograms) taken by the Solar Optical Telescope on board Hinode. We have generalized the model equation of image evolution by adding to the continuity equation a source term that is proportional to the image value. Synthetic images were constructed based on the analytical solution of this equation with different velocity profiles: uniform, affine, or nonaffine. The tests with the synthetic data indicated that NAVE is very good at detecting subpixel motions, superpixel motions, and nonuniform motions, while LCT is not good at detecting nonuniform motions, especially around critical points, and the performance of DAVE is degraded in the presence of superpixel motions. In all the methods, the performance became worse as the velocity field deviated more from an affine one. We also found that the MHD simulation data we used are not quite suited for discriminating between the three methods, maybe because the data do not contain enough structural information to be used for tracing. In contrast, the determination of velocity fields from the real image data was somewhat sensitive to the technique adopted. The technique of NAVE with the source term produced velocity fields that are the most consistent with the data.
- 出版日期2008-12-10
- 单位中国科学院国家天文台
