应用空间加权方法的二维图象运动估计研究

Spatio-Weighting-Function Scheme for the Motion Estimation of 2-D Image

图象运动估计是计算机视觉及智能图象制导研究领域中的一个重要课题．传统的微分法需要求图象的时空导数，在实际计算中用差分代替微分使微分法的抗干扰能力差．本文采用图象空间加权方法研究二维困象运动估计，书求图象的空间导数转化为对加权函数求导，从而提高了抗干扰能力；此外，空间加权函数在满足一定条件下可任意选择，并且这种方法可自动适应图家的大小变化．实验表明，本文的运动估计算法精度高，抗干扰能力强．

One of the most important Problems in image sequence analysis is motion estimation. Considering real-time motion detection and multipk target tracking, the author presents an alternative approach for recovering the target velocity by introducing six spatio weighting functions. The advantages of this scheme are: (1) target velocity is derived directly, without computing local velocity at a number of image locations; (2) the scheme requires computaton of first derivatives rather than second derivatives, and is more resistant to noise as compared with differential method; (3) an appropriately large image patch is selected for the computation of velocity of camera relative to background, but for the computation of target velocity, an appropriately small image patch is selected; (4) the scheme is well suited to implementation in parallel architectures (digital or analog). TV tests show that the author's scheme gives real-time estimahon results of good precision.