基于SIFT特征的视频稳像算法

为将载体上截取的视频进行稳像,介绍了尺度不变特征变换(scale invariant feature transform,SIFT)算法及基于SIFT特征的视频稳像算法.基于电子稳像算法框架采用六参数仿射模型表征图像运动,利用图像的SIFT特征估计图像间的模型参数,通过补偿来消除图像间的运动,完成了视频稳像任务.最后,利用峰值信噪比(peak signal to noise ratio,PSNR)指标评价稳像效果.实验结果表明:该算法能很好地消除图像间的运动,有效提高图像间的PSNR值,完成视频序列的稳像任务.

Video Stabilization Algorithm Based on SIFT Feature

In order to stabilize the video obtained from cameras on kinds of carriers, this paper introduces the scale invariant feature transform (SIFT) algorithm and video stabilization algorithm based on SIFT features. The motions between images are represented with the model of affine model with six parameters, which is estimated by using the SIFT features of images. Then the images are compensated so that the motions would be counteracted and the goal of stabilizing the video would be achieved. At last, peak signal to noise ratio (PSNR) is imported to evaluate the effect of the algorithm. As the experiments show, algorithm proposed in this paper can counteract the motions among images very well and can improve the values of PSNR effectively. Video stabilization of video sequence is also completed with the algorithm.