动态前景干扰下的电子稳像算法

Electronic Image Stabilization Approach under Dynamic Foreground Interference

摘要:

针对存在动态前景目标的抖动视频序列, 提出了一种以块匹配为基础的块差分平方前景目标去除的电子稳像方法。首先, 将图像分成大小为16×16的特征块, 通过设定阈值筛选出具有足够梯度信息来估计运动矢量的特征块, 从而减少参与计算的特征块数量以提高处理速度; 然后, 采用归一化互相关(Normalized Cross Correlation, NCC)匹配算法和迭代搜索策略来完成特征块的匹配; 最后, 采用牛顿迭代法计算仿射变换模型下的帧间运动参数, 并根据这些参数完成运动补偿, 实现图像序列的稳定。实验表明, 算法成功地去除了视频序列中动态前景目标的干扰, 对存在平移及旋转运动的视频序列, 能够快速有效地去除抖动, 获得稳定的视频序列。

Abstract:

Aiming at the shakes of video sequences with dynamic foreground targets, an electronic image stabilization approach using block difference square based on block matching was proposed to remove the foreground moving target. First, the image was divided into feature blocks with the area of 16×16. Feature blocks with enough gradient information were filtered out by setting threshold to estimate the motion vector, and then the number of feature blocks was reduced to enhance the processing speed. Then, normalized cross correlation matching algorithm and iterative searching strategy were used to execute the matching of feature blocks. Finally, Newton iteration method was adopted to calculate the interframe motion parameter under affine transformation model, and motion compensation was accomplished by means of these motion parameters, then stabilized image sequences were output. Experimental results demonstrate that the proposed approach can remove the interference of foreground moving target, and can eliminate the jitter of image sequences in the form of translation motion and rotation motion fast and efficiently, and stable video sequences were achieved eventually.