不同场景下虚拟现实视频畸变校正算法

Distortion Correction Algorithm of Virtual Reality Video in Different Scenes

不同场景下虚拟现实重构视频受到运动目标以及载体方位不稳定等因素的影响导致视频畸变,提出基于运动补偿及特征图像配准的虚拟现实视频畸变校正算法.基于兴趣特征点检测的方法建立不同场景下虚拟现实视频图像检测模型,通过图像配准和角点稀疏性检测,实现对虚拟现实视频进行畸变成分检测、动态描述和特征匹配,将不同场景下虚拟现实视频的图像帧投影到一个公用坐标系,并从相邻帧的背景层中分离出视频的畸变成分,采用运动补偿及特征图像配准的方法实现视频畸变动态校正.仿真结果表明,采用该方法进行不同场景下虚拟现实视频畸变校正的动态配准能力较好,适合不同复杂场景下的视频图像虚拟现实重构和检测.

In different scenes,the reconstructed video of virtual reality is distorted by the images of moving objects and the unstable orientation of the carrier.This paper proposes a distortion correction algorithm of virtual reality video based on motion compensation and feature image registration.Based on the method of interest feature point detection,the image detection model of virtual reality video in different scenes is established.Through image registration and corner sparsity detection,the distortion component detection,dynamic description and feature matching of virtual reality video are realized.The image frames of virtual reality video in different scenes are projected to a common coordinate system,and the distortion components of video are separated from the background layers of adjacent frames.Motion compensation and feature image registration are used to realize the dynamic correction of video distortion.The simulation results show that the dynamic registration ability of virtual reality video distortion correction in different scenes is good,and it is suitable for video image virtual reality reconstruction and detection in different complex scenes.