形状与内容保护的多摄像机视频融合方法

Shape and content preserving multiple cameras video fusion method

视频融合变换模型的设计是视频融合的关键技术,传统的视频融合变换模型使用单一的投影变换矩阵,投影变换在视频融合过程中通常会引起视频帧图像的变形,尤其是对于存在较大视差的视频帧图像,变形会更加严重.针对此问题,提出一种形状与内容保护的多摄像机视频融合变换模型,该模型通过拼接缝查找算法得到接缝均值,使用接缝均值将图像划分为两个半空间,在左半空间进行投影变换,右半空间使用相似变换代替投影变换,投影变换与相似变换的结合保护了视频帧图像在变形过程中不会发生较大的拉伸与缩放;同时在拼接缝邻域应用内容保护变形算法,更好地保证了过渡区域的融合精度.把新的视频融合变换模型应用于实际拍摄的视频数据上,实验结果表明该方法具有较高的鲁棒性与较强的实用性.

The design of video fusion transform model is an important step in video fusion.Projection transformation matrix is used in the current multiple cameras video fusion transform model,but it usually causes the severe distortion.When there are larger parallax among cameras,the distortion will be larger.So the traditional video fusion transform model has a higher request for the camera position.To solve this problem,this paper presents a shape and content preserving video fusion transformation model.Finding the seam and getting the similarity transformation matrix are the keys of the proposed model.The algorithm of dynamic programming is used to find the seam,and the similarity transformation matrix is calculated according to the continuity between the projection transformation and the similarity transformation.When we get the seam,the space is divided into two half spaces.Projection transformation model is used in the left space,and similarity transformation model is used in the right space.The combination of projection transformation and similarity transformation protects the shape does not occur with larger tensile scaling.In addition,content preserving algorithm is used in the seam neighborhood,which causes a better result.Applying this new model to the actual video data,the experimental results show that this method has high robustness and practicality.