运动估计是电子稳像算法的主要步骤,其运算复杂度直接影响稳像的实时性,为此提出了一种利用灰度码位平面和搜索模板进行运动估计的电子稳像方法.该方法首先计算出视频序列帧图像融合相邻位平面信息的二值图像,然后通过蜂窝状搜索模板和小菱形搜索模板计算出当前帧和参考帧图像的最佳运动矢量位置,最后进行运动补偿实现视频稳定.实验结果表明,该方法能够在不影响稳像精度的前提下,显著降低计算复杂度.%Motion estimation is the main step of image stabilization, which directly affects the computational complexity of image stabilization. To reduce the computational cost,a method for electronic image stabilization based on gray bit-planes motion estimation is proposed. Firstly, the binary image with fused bit-plane information of adjacent video sequences frames is calculated. Secondly, the best motion vector position for the current frame and reference frame image is searched by honeycomb and small diamond search pattern. Finally,motion compensation is complemented to obtain the stabilized video frame. Experimental results show that the computational complexity of image stabilization is greatly reduced, while the accuracy of image stabilization is not affected obviously.
展开▼