云雾条件下视频稳像的预处理

Preprocessing on Foggy and Cloudy Video Stabilization

受云雾干扰的视频图像清晰度差,对比度低,提取的特征点正确率低,影响运动参数的计算,导致稳像精度下降甚至稳像失败。采用改进的二维OTSU算法进行分割预处理,获取图像中较清晰的区域,得到正确率较高的特征点。针对不同的情况采用图像增强和多次分割进行优化,再通过八邻域法剔除孤立点。实验表明,该算法实时好,能适用于不同云雾条件下的视频图像,分割得到的目标区域能在提取足够数目特征点的同时保证特征点的正确率,为后续电子稳像打下坚实的基础。

The video images affected by the fog and cloud are always misty and low-contrast. This can make the accuracy of the feature point low, which impacts the motion parameter and results in the low precision, even the failure of the image stabilization. The improved two-dimensional OTSU algorithm is proposed in this paper, which is used to gain the clear area in the image, so that the feature points can be detected correctly during the video stabilization. Image enhancement and multiple segmentation are used to improve the algorithm, which make it more effective in different conditions. Therefore, eight-neighborhood method is used to eliminate outlier. Experimental results show that the algorithm above can be applied to the different conditions, which can be used to obtain enough feature points correctly in real time. The solid foundation is laid for the video stabilization.