基于GPU的视频流拼接算法研究

为解决视频流的稳定实时拼接,结合图形处理器GPU强大的并行计算能力,提出了一种基于GPU的视频流拼接算法.提取视频流的帧图像,利用尺度不变特征变换(scale invariant feature transform,SIFT)算法在GPU上实现帧图像的特征提取与匹配,实现图像拼接,进而实现视频流的稳定实时拼接.基于GPU的SIFT算法充分利用了GPU的并行处理能力,加快了视频流拼接算法执行的速度,真正意义上实现了几个差异较大但具有公共视野的视频流快速稳定的拼接.

Research of video stream splicing algorithm based on GPU

To solve the stability and real-time of video stream splicing,combined with the powerful graphics processor GPU’s parallel computing capabilities,a design method of stream splicing algorithm based on GPU is presented.Extracting video stream frame image,then image stitching which contains feature extraction and matching is implemented on the GPU using SIFT(scale invariant feature transform) algorithm,to realize the stable and real-time video stream splicing.The SIFT algorithm based on the GPU makes full use of the GPU’s parallel processing capability,which accelerates the implementation of video streaming stitching algorithm and realizes the fast and stable video stream splicing with quite different but a public vision.