基于四参数仿射模型的频域运动估计技术

仿射模型能比块平移模型更好地拟合真实运动场景,但待解的参数也相应地增加了.基于梯度的搜索算法在高阶问题中存在局部极小和收敛速度的缺陷,基于Hough变换和Gabor滤波等的变换域方法同样存在运算量的问题,影响了仿射模型在实时运动估计中的应用.传统的相位相关法仅利用了傅氏频谱的相位信息求解块平移而未利用频谱的幅值特性.本文通过对仿射运动场频域特性的分析,将仿射模型的线性项和平移项解耦,结合频谱的幅值特性和相位特性在频率域内求解仿射运动参数.实验结果表明该算法在运算量增加不大的情况下运动估计效果得到明显改善.

Two New Magnetic Localization and Parameter Estimation Methods Under the Dipole Model

In the paper,analyses the time and frequency domain characteristics of the target signal under the classical magnetic dipole model and points out the difficulty and particularity of the magnetic localization and parameter estimation problem.Designs the second order difference matrix method (SDMM) and frequency domain matching method (FDMM) to solve the localization and parameter estimation problem.Both of the algorithms make use of the time and frequency domain characteristic of the signal and convert a high order optimization problem into sequential solutions of several low order problems.SDMM has low computation burden,excellent real time performance.FDMM is more reliable in low SNR environment and suitable for post analysis.Simulation of both methods meets well with the theoretical analysis and the algorithms achieve the goal.The paper also points out the further work for the two algorithms.