Projection And Coding Optimization Algorithm Of Omnidirectional Video

Recently,Virtual Reality(VR)and Augmented Reality(AR)technologies have attracted a large number of users due to their rich immersion and have a wide range of applications in people's lives.Essentially,the core of VR and AR technologies is omnidirectional video,which is usually represented in a spherical domain.Its high frame rate and high resolution requirements have promoted the explosive growth of video data volume,which has brought severe challenges to storage and transmission.To compress omnidirectional video effectively,the JVET organization jointly established by the International Telecommunication Union-Telecommunication standardization sec-tor(ITU-T)and the International Organization for Standardization(ISO)/International Electrotechnical Commission(IEC)has specially developed a 360 Library(360Lib),which can be combined with High Efficiency Video Coding(HEVC)or Versatile Video Coding(VVC).The framework of omnidirectional video coding mainly contains two parts,i.e.,projection and coding.There are oversampling and discontinuous boundary problems in the projection part.In the coding part,the traditional coding standards do not fully consider the spherical characteristics of the omnidirectional video.Based on the content characteristics,texture characteristics and neighborhood characteristics of omnidirectional video,this paper has carried out the following works:1.The fixed mapping functions mostly used in existing projection formats cannot be applied to diverse content.In this paper,considering the content characteristics of omnidirectional video,a content-aware Hybrid Equal-angular Cubemap(CHEC)pro-jection format is proposed.Firstly,a new content-aware projection model is proposed,in which the parameter of mapping function can be adapted to different video con-tent.Secondly,by minimizing the projection distortion,an iterative parameter search algorithm is used to determine the optimal parameter of mapping function for each projected plane.Finally,the projection model is incorporated into the framework of omnidirectional video coding to improve coding performance.Compared with the Equi-Rectangular Projection(ERP)and Hybrid Equi-angular Cubemap(HEC),the proposed CHEC projection format has achieved 8.57%and 0.11%bit savings respectively on average under the end-to-end Weighted to Spherically uniform Peak Signal-to-Noise Ratio(WS-PSNR),which effectively improves coding performance2.Oversampling and discontinuous boundaries are the main factors causing pro-jection distortion.In this paper,a Spherical Rotation-based Omnidirectional Video Coding(SROVC)method is proposed.Firstly,a framework of SROVC is proposed to improve coding performance,in which Full traversal-based Spherical Rotation(FSR)adopts frame-level rate distortion optimization to determine the optimal rotation angle Secondly,a Texture characteristics-based Spherical Rotation(TSR)is proposed to pre-dict the optimal rotation angle based on texture characteristics,which achieves coding gains equivalent to that of FSR method with less coding times.Finally,a Group-oriented Texture characteristics-based Spherical rotation(G-TSR)is proposed to maintain coding gains and reduce complexity,where the number of groups is determined by statistical experiments.The proposed TSR and G-TSR methods achieve an average bit saving of 0.94%and 0.89%respectively for the CMP format3.The existing single intra mode cannot adapt to diverse content changes,and the additional coded mode index in the combined mode costs more bit overhead.In this paper,a combined weighting-based intra prediction method is proposed to improve coding performance.Specifically,the combined modes are obtained from the first two modes in the Most Probable Modes(MPM)list,which do not need to be transmitted to the decoder side through encoding.Additionally,the weight value of the combined weighting needs to be manually set and do not need to be transmitted to the decoding side The proposed intra prediction method based on MPM weighting is incorporated to the traditional video coding framework and the omnidirectional video coding framework,and the rate-distortion optimization is used to compete between the traditional intra prediction method and the MPM weighting intra prediction method.Compared with traditional video standards,the proposed algorithm achieves an average coding bit saving of 0.02%for 2D imagesThe projection and coding optimization algorithm of omnidirectional video is proposed in this paper.It adds the mapping function parameters that combine the content characteristics to the projection for reducing the projection distortion.The spherical rotation model considering the texture characteristics is added to the framework of omnidirectional video coding for improving coding gain.The combined weighting intra prediction model combined with neighborhood information is icorporated into intra prediction for reducing spatial redundancy.The research achievements obtained in this paper can be applied in military,education,medical,entertainment and other fields,and are of great significance to the development of digital media and Internet industries.