【摘要】 运动分析是视觉信息加工中的一个重要问题。本文利用Ｒｅｉｃｈａｒｄｔ的相关型初级运动检测器（ＥＭＤ）二维阵列可以有效地进行图象－背景相对运动分辨的特点，以及小波变换的频谱分析特性与人类视觉多频率通道特性相类似的性质，将ＥＭＤ模型、小波变换和图象的塔式结构处理有机地结合起来，提出了一种类似视觉信息加工方式的新的运动分析算法。计算机仿真结果表明，算法能够较好地模拟视觉运动检测的功能，与Ｈｏｒｎ＆Ｓｃｈｕｎｃｋ算法［１］相比，提高了运动估计的速度与精度。更多还原

【Abstract】 Motion analysis plays an important role in visual information processing. It has been proved that Reichardt’s Elementary Motion Detector (EMD) can perform figure-ground discrimination efficiently, and Wavelet transform analysis is similar to the multi-frequency channel of human vision. So we combine EMD, Wavelet transform and Pyramid image processing together, to propose a new motion estimation algorithm. The computer simulation results reveal that the proposed algorithm can well simulate vistual motion evaluation. Compared to Horn & Schunck’s algorithm, this algorithm provides more effective and accurate motion estimation.更多还原