Research On Stitching Quality Assessment Method Of Virtual Reality Panoramic Video

With the rapid development of virtual reality technology and its wide application in various fields,VR panoramic video got widely development with its immersive and interactive experience.Currently in virtual reality platform content,VR panoramic video content occupies very important position.VR panoramic video which is based on image mosaic algorithm is the most widely used,and the quality of video stitching has a direct impact on the quality of panoramic video.Now stitching algorithm is various,so the evaluation of the quality of video stitching not only can provide the basis for optiming panoramic video stitching algorithm,but also can inprovide the panoramic video immersive and realistic factors of quantitative evaluation.Panoramic video is the main form of virtual reality technology,which uses multiple cameras and mosaic generation can get around 360 degrees on the scene at the same time,truly show around the virtual environment.However,because of the human visual system and the limitted camera filming equipment,we cannot directly get 360 degrees panoramic images.It needs a professional virtual reality system acquisition device or using image stitching algorithm to stitch the overlapping area of video frames got from multiple cameras.At present,most of panoramic video stitched from overlap area of the video stream using video stitching technology,but due to different application scenarios of video leading to mosaic method will be different,the video quality will also have certain differences.In this paper,based on the study of panoramic video mosaic algorithm,we evaluate the quality of VR panoramic video stitching and optimize the quality of panoramic video stitching.This paper analyzes the key technology in the process of VR panoramic video mosaic:panoramic image mosaic and image fusion,main factors affecting the quality of VR panoramic video mosaics,proposed the quantitative evaluation algorithm of VR panoramic video mosaic quality using image feature information.The influence of different factors on the quality of video frames is analyzed,and the corresponding image features are extracted.So we puts forward to quantitative evaluation algorithm by using image feature information of panoramic video for stitching quality.Analysis different factors impact on the stitching quality of video frames and get the corresponding video frame characteristics.Evaluation model based on BP neural network is established.Then subjectively assessment stitched video with different qualities and map image feature parameters to subjective evaluation scores using BP neural network.Finally training the model of BP neural network and getting objective evaluation results of VR panoramic video stitching quality by trained model.This algorithm has some reference value for the evaluation of the quality of VR panoramic video mosaic.