一种基于灰度投影与块匹配的视频序列快速稳像算法

摘要

针对视频图像序列的不稳定性,结合灰度投影算法与块匹配法各自的特点,提出了一种可以快速消除图像序列间平移和旋转抖动的新型电子稳像算法.该算法首先采用灰度投影法估计与补偿图像序列间的平移运动,然后用拉普拉斯算子在图像边缘处选取灰度信息比较丰富的小块模板,利用块匹配法计算模板的平移运动,再结合模板在图像上的位置估计图像的旋转运动,最后通过运动补偿得到稳定的视频图像序列.实验结果表明,该算法可以明显地减轻图像序列间的抖动现象,对于提高动态图像的稳定性具有很好的效果.%In allusion to the instability of video sequences, a new electronic image stabilization algorithm, based on different features of gray projection algorithm and block matching algorithm, is proposed to fast eliminate the shift,dithering and rotation of video sequences. This algorithm first adopts gray projection algorithm to reckon and compensate the translational motion between video sequences, and then adopts Laplace operator to select small templates with abundant gray-scale information near the edge of the image. Next, it uses block matching algorithm to calculate the translational motion of templates. Moreover, it reckons the image rotation according to where the templates locate on the image. In the end, it uses motion compensation to attain stable video sequences. Experimental results show that, the algorithm can obviously eliminate the dithering between video sequences, which will have a salutary effect on improving the stability of dynamic image.