An objective quality metric for panoramic videos

The creation of high quality panoramic videos for immersive VR content is commonly done using a rig with multiple cameras covering the required scene. Unfortunately, this setup introduces both spatial and temporal artifacts due to the difference in optical centers as well as the imperfect synchronization between them. Therefore, designing quality metrics to assess those videos is becoming increasingly important. Using traditional image quality metrics is not directly applicable due to the lack of a reference image. In addition, such metrics do not capture the geometric nature of such deformations. In this paper, we present a quality metric for panoramic video frames which works by computing pair-wise quality maps prior to blending and fusing them to obtain a global map of potential errors. Our metric is based on an existing one designed for novel view synthesized image, which is a similar problem to image stitching. Results show that applying this quality metric offers a practical way to assess panoramic video frames which do not have a reference. It also consolidates the similarity of the artifacts produced from novel view synthesis algorithms and those produced in the process of image and video stitching