Generalized Cubemap Projection For 360 VR Videos

360 VR videos record a surrounding environment in every direction at the same time and enable viewers to interactively look around in the captured environment,providing the audience with a 360 degree immersive experience.Compared to computer-modelled 3D content,360 VR videos can capture immersive content very quickly.Due to this unique feature,it opens a new world of potential applications such as VR education/training,VR games,VR films,VR museum/tour and VR sports broadcasting.360 VR videos face a few new challenges.First,the bandwidth demands of streaming high quality videos are vastly increased compared with conventional planar videos.Second,the 360 VR video storing and editing process is in a way similar,but also very different from conventional videos.To encode 360 VR videos,spherical pixels must be mapped onto a two-dimensional domain to take advantage of the existing encoding and storage standards that are designed for conventional videos.In VR industry,Cubemap Projection(CMP)is the most widely used projection method for encoding 360 VR videos.It deforms a sphere into a cube,then unfolds the cube's six faces and packs them into a flat rectangle.However,it exhibits pixel density variation at different regions due to severe projection distortion.In this dissertation,we present a new projection algorithm to tackle the problem mentioned above.More specifically,·We present a generalized CMP algorithm to improve the efficiency of standard CMP using polynomial approximation.·Based on the proposed generalized CMP,we perform error analysis in terms of local distortion,pixel density uniformity.·We use PSNR to evaluate and compare the visual quality.We conduct intensive experiments and the results demonstrate promising performance improvement against the state-of-the-art,such as standard ERP,CMP,EAC and some other projections.