基于相机姿势全局运动估计及改进修复算法的视频稳定技术

Video Stabilization Based on the Camera Pose a Global Motion Estimation and Improver Restoration Algorithm

在视频稳定的过程中,由于摄像机的运动,造成图像的扭曲。针对这种情况,提出一种基于相机姿势的全局运动估计,同时为了克服图像拼接后,部分区域像素丢失的问题,使用改进后调和模型来修复缺少的像素。算法首先提取特征不变量,然后基于这些特征不变量去估计摄像机的运动矢量,相乘各帧间的运动矢量,可以得到每一帧参考第一帧的运动矢量。运用这个矢量可以很好地计算出没有扭曲的图像。运用计算出的图像与视频帧进行拼接,可以很好的解决图像的扭曲的问题。然而,图像拼接完成后可能导致部分区域像素缺少,为了填充缺少像素,算法使用了改进的调和模型来修复缺少区域。实验结果表明,基于相机姿势的全局运动估计可以很好的解决图像扭曲的问题,同时改进的调和模型可以高效的完成对图像的修复。

In the process of video stability,due to the movement of the camera,the image distortion.To solve this problem,this paper offer a method of global motion estimation based on the camera position,at the same time in order to overcome the problem of the pixef leakage after the image stitching,using the improved harmonic model to repair the lack of pixels.Extracted feature invariant algorithm firstly,then based on these characteristics to estimate the camera motion vector,multiply the motion vector of each frame,each frame can be obtained refer to the first frame motion vector.Using the vector can be good to calculate don＇t distorted images.Used to calculate the images and video frames can well solve the problem of image distortion.