| With the rapid development of multimedia technology,Virtual Reality(VR)is popular and commonly used for its immersive user experience.Most video information in VR is spherical video.Generally,spherical video has a pretty high resolution requirement due to user experience.Without compression,the amount of data is huge,which bring a great challenge to ordinary hard disk storage and network transmission.Therefore,it is very important to improve the coding efficiency of spherical video.Due to the limitations of the existing video coding standard,the spherical video needs to be projected into the two-dimensional panoramic video before being transmitted to the codec system.At the receiving terminal,the two-dimensional panoramic video is converted into a spherical video for users to watch.As a result,it is more reasonable to make the spherical video distortion as the distortion measurement of panoramic video coding.The common spherical video distortion evaluation index at this stage are spherical PSNR,weighted to spherically uniform PSNR,craster parabolic projection PSNR.Based on this,we take the above mentioned spherical video distortion as the measurements of the quality of the viedo to optimize the panoramic video coding technology.Equi-rectangular projection,cubemap projection,octahedron projection,segmented sphere projection is the typical projection format of panoramic video.Different projection formats have different characteristics due to different projection algorithms.However,all of the four projection formats have interpolation operations,which change the equal-weight characteristics of spherical pixels.Since the existing encoder is still used in panoramic video coding process,rate-distortion optimization is used during the mode selection process.In this paper,the rate distortion optimization technique based on the measurement of spherical distortion is proposed.The relative weights of the different projection formats are introduced to change the proportion of the distortion in the region of different interpolation conditions during the calculation of the rate distortion cost.So as to achieve the equal-weight characteristics of spherical pixels and improve the coding efficiency of panoramic video.The proposed rate distortion model can save 1.47 percent coding rate on average.In spite of the fact that equi-rectangular projection is the most accepted panoramic video projection format,the inter-frame prediction of the coding process in ERP format is optimized.Due to the high resolution requirement of panoramic video,the local advanced motion vector resolution is introduced into High Efficiency Video Coding,to enhance the quality of high definition panoramic video.In this paper,the present coding block is prejudged by a proposed texture complexity measurement and the appropriate threshold is selected to optimize the high latitude region of ERP format while using LAMVR technology.Simultaneously,according to the current coding block interpolation condition of the high latitude area for ERP format,the sub-pixel accuracy motion estimation is optimized after the texture complexity determination is adopted,in an effort to keep the video quality at the same and reduce the encoding time complexity.Search algorithm of motion estimation is also optimized in inter prediction based on ERP format.The proposed scheme can reduce the coding time by 5.57 percent on average while have the same performance with the LAMVR technology.The time saving of the proposed scheme is particularly obvious with the sequences that have global movements or large amount of motion,with the maximum coding time being reduced by 10.53 percent. |
PDF Full Text Request