Research On Inter-frame Error Concealment For 3D Video

The 3D video technology can not only provide the audience with an immersive immersion but also present the user with different angles of video images,and enhance the user's three-dimensional immersion and scene fidelity.It is an important development direction of multimedia technology.A large amount of data of 3D video brings massive pressure to the transmission channel,and it must be efficiently compressed and transmitted.However,highly compressed bitstreams are more susceptible to many uncertainties in the network during transmissions,such as network delay or noise interference.If the bitstream loses during transmission,the decoded image will be distorted.In severe cases,the entire frame of the image may be missing,which may cause the subsequent decoding of a series of frames to fail.In order to solve such problems,reliable error control techniques must be employed.The 3D video error concealment technique uses the time-space domain and Inter-views information included in the 3D video to reconstruct the lost data at the decoder.How to obtain relevant information from the inter-view frames,inter frames,and an intra frame included in the 3D video,and apply it to reconstruct the error region is the core problem in the 3D video error concealment algorithm.This thesis analyzes the possible correlation between the current error frame and other frames.It focuses on reconstructing the distorted video frames based on the redundant information coming from the spatial,temporal,and inter-views of the 3D video stream.The main work of this thesis is listed as follows:1.We propose a 3D video error concealment algorithm based on the HEVC standard.Based on the HBP structure,the algorithm determines the region type of the missing block by using the pixel difference between the forward block and the backward block.For static blocks,a fast and efficient zero vector copy method is used to restore them.For the motion blocks,the algorithm uses the human eye perception to segment the missing block and restore the lost region within the same perception range.The LBP operator is used to detect the texture region of the initial recovery block,and the quality of the missing block whose boundary texture is inconsistent is further improved.The experimental results show that the algorithm effectively improves the reconstruction quality of the lost blocks in the 3D video and maintains the texture consistency of the restored blocks in different test sequences.2.We propose a 3D video error concealment algorithm using the coding characteristics of blocks.The traditional 3D video error concealment techniques generally recover all lost blocks using a single method,without taking into account differences in different missing blocks.Based on the statistical analysis of the coding prediction mode of the 3D video sequence coding block,the algorithm predicts the coding prediction mode of the current lost block and divides it into three types of coding blocks.According to the coding mode of the missing block,different error concealment strategies are applied.In the case of inter-view coding block repairment,the gray level co-occurrence matrix improves the traditional boundary matching criterion,which improves the matching accuracy between the lost block and the multi-candidate block.The experimental results show that the repaired lost blocks are closer to the original information,which can significantly enhance the subjective and objective quality of the reconstructed video.