Saliency Detection Of Panoramic Video And Its Application In Coding And Transmission

In recent years,the content presentation of panoramic video has been widely concerned by industry and academia.Panoramic video has the characteristics of arbitrary switching angle,high resolution and high bitrate.It provides users with immersive visual experience without blind areas.Because of the large amount of panoramic video data,it also causes some new problems in the process of coding and transmission:(1)the angle of view after switching often does not have a clear view,resulting in uncomfortable user experience;(2)the network transmission bandwidth restricts the transmission of high definition video;(3)the traditional coding and transmission schemes do not take into account the panoramic video characteristics.Therefore,in order to improve the visual experience of panoramic video more efficiently,this paper combines the saliency detection with the optimization coding and transmission research.First,we study the saliency detection algorithm of panorama.Then the detection results are applied in panoramic video optimization encoding and transmission,to provide similar video senses for users and effectively reduce the amount of coded data and transmission bandwidth.Firstly,panoramic video has a wider range of vision and more complex scenes.How to accurately determine the area of interest in human eyes and to further improve the user's visual experience in encoding and transmission is a solid foundation.Therefore,the saliency detection of panorama is the most important step.Considering that panorama has no boundary,this paper proposes a saliency detection algorithm for panorama based on multi-angle segmentation,which cuts panorama from different angles to different versions of the same scene.Then,based on the cube projection,we use an improved dense and sparse reconstruction algorithm to detect each surface for saliency.Last,the original panorama is obtained by reverse mapping and multi-angle fusion,and the best segmentation times are verified.Secondly,based on the principle of high quality coding for the viewing area,we add the weight of the motion vector as the visual characteristics of the human eye based on the saliency value.Subsequently,we divide the viewport area and out of the viewport area by different quantization parameters.The coding of panoramic video can also reduce the bitrate while improving the user experience.Experiments evaluates the effectiveness of this coding scheme objectively and subjectively.Finally,we build the adaptive transmission system,and study the low delay solutions in transmission from time and space.After switching the angle of view,the user needs to request the high quality video slice version in the new angle of view.Due to the dependence of the video stream,the angle of view after the switching must wait until the arrival of the next I frame,which is bound to bring a certain delay.The tile division on the space is too large,making the section of the video slice unavailable after switching the angle of view,wasting the bandwidth and increasing the delay.Therefore,in space,the tile of different sizes is divided according to the significant weight,and it is proved that the significance is important to reduce the delay in the transmission.