Research On 360-degree Panoramic Video Of Fast Encoding Method For The Versatile Video Coding Standard VVC

Nowadays,virtual reality technology is becoming more and more mature.As a novel content carrier that provides an immersive experience,360-degree panoramic video has been widely spread,and it has gradually become a research hotspot in computer science,military,industrial and commercial applications.Compared with traditional video,360-degree panoramic video scene coverage is wider and the amount of data is larger,but at the same time,th e problem is that the transmission is more time-consuming,it takes up more storage space and transmission bandwidth,and requires more equipment hardware.Higher,so the coding efficiency problem needs to be solved urgently.This thesis aims at the problem of long encoding time caused by the high complexity of 360-degree panoramic video encoding under the universal video coding standard H.266/VVC(Versatile Video Coding,VVC),and explores ways to reduce encoding time while ensuring certain encoding perfor mance.The main research contents are organized as follows:First of all,in the 360-degree panoramic video coding process,the coding time is too long due to the high coding complexity of the block division method decision process,this paper proposes a fast block division method decision algorithm.This algorithm first uses the equidistant rectangle(Equirectangular Projection,ERP)mapping format to divide the entire graph by the uneven sampling characteristics at the poles and the equator,and initially reduces the coding tree unit(Coding Unit,CU)where the current coding unit(CU)is located.Coding Tree Unit(CTU)coded depth interval.Then use the imported AC coefficient energy to judge the complexity of the image texture of the CTU,and further limit the coded depth interval of the current CTU.Then use the spatial correlation of the CTU to further reduce the coded depth interval of the current CTU.Finally,the coded depth prediction of the current CU and the early termination of the division method is realized by using the coded depth interval of the current CTU.This improvement greatly reduces the calculation amount of the rate-distortion cost(Rate Distortion Cost,RDC)under the corresponding block division method.As experimental results showed,the proposed algorithm achieves an average of 44.06% encoding time savings than the original algorithm,and at the same time,it causes less bit rate increase and lower peak signal-to-noise ratio drop.Secondly,in view of the time-consuming problem of coding mode decision-making process in 360-degree panoramic video coding,a fast algorithm in mode decision-making is proposed in this paper.This algorithm first uses the correlation of coding modes between adjacent coded depths to initially filter down the number of coding modes.Then use the multiple relationship s of Hadamard's change generation value to further reduce the number of coding modes.This algorithm achieves fast coding by reducing the number of coding modes and then reducing the amount of calculation of the cost of the corresponding coding mode.Compared with the original algorithm of the VVC standard,the average coding time of the proposed algorithm is shortened by 27.63%,while the increas e of the code rate and the decrease of the peak signal-to-noise ratio is not obvious.In addition,since both the block division method decision process and the coding mode decision process belong to the intra-frame prediction process,this article also combines the two algorithms to generate a fast intra-frame prediction algorithm.Under the objective evaluation index,compared with the original implementation of anchor,the average coding time of the proposed algorithm is reduced by 55.77%,and the coding performance has a certain degree of degradation.Under the subjective evaluation index,the image distortion produced by the proposed algorithm is negligible.Finally,in view of a large amount of calculation in the inter-frame prediction process in 360-degree panoramic video encoding,this paper proposes a fast inter-frame prediction algorithm.The algorithm first uses the temporally encoded adjacent frames to determine the motion complexity threshold,and then uses the motion complexity of the spatially adjacent CUs that have been encoded to predict the intensity of the current CU,and then target different intense motions.Different motion search range reduction strategies are formulated to reduce the amount of calculation of rate-distortion cost calculation when the current CU is looking for the best motion vector.It can be seen that,compared with the original encoder,the proposed algorithm achieves an average coding time saving of 0.88% with little performance degradation.