| With continuous development and wide application of digital video technology,the traditional 2D video has become increasingly difficult to meet the needs of users.Therefore,people now mainly focus on the research of multiview,free viewpoint and 3D video technology.The expression of multiview texture plus depth image has the characteristics of low complexity,good effect of virtual viewpoint rendering,and it is compatible with2 D video equipment,which makes it has broad application prospects.At the same time,multiview video data is too large to compress,store and transmit,which is also the difficulty of multiview video technology.Therefore,under the premise of limited hardware and channel capacity,video coding rate control becomes very important.Based on the principle of coding optimization,the following three aspects are studied in this paper:(1)Research on the fuzzy logic and fuzzy control theory and algorithm.Combining with the fuzzy control theory and multiview video coding structure,constructing a fuzzy controller which is suitable for multiview video coding rate control,and basing on the relationship between the coding buffer errors and the coding parameters,the fuzzy control model is established.A multiview video coding rate control algorithm based on fuzzy logic is proposed,which can achieve a good control precision and improve coding performance.(2)Research on multiview video coding time domain rate distortion optimization.Based on the theory of rate distortion,a method of calculating the interframe time domain propagation chain under multiview video coding is proposed by studying the time domain propagation chain.The time domain distortion propagation model is constructed by establishing multi interframe time domain propagation chains for multiview,and introducing the Lagrangian multiplier in coding control.Based on the study above,a multiview time domain rate distortion optimization algorithm is proposed.(3)Research on video image background modeling.In order to solve the problems of coding and storing of massive surveillance data,a block-based background modeling algorithm is proposed.It uses the replacement method based on variable blocks,and makes the classification detection and selection according to calculate the residual and gradient in the block,and makes a further selection by the block boundary detection method.In addition,based on the background modeling algorithm,a long term reference algorithm and adaptive training set algorithm are proposed which can maximize its performance.Combining with the research of the three aspects above,this paper proposes a common multiview video coding rate control system which can be directly applied to the current video coding standards.The research on rate control and rate distortion optimization has been integrated into the AVS 3D video coding reference software RFD.As a part of the standard,it achieves rate control and rate distortion optimization on the multiview and 3D profiles. |
PDF Full Text Request