Research On RGBD Video Sequence Preprocessing And Quantization Encoding

In recent years,the depth camera has made great progress in the field of consumer electronics and industrial automation.The processing and compression of the depth map has become a hot topic in the academic community.In the paper,with the support of the deep camera compression coding project,the depth map recovery technology and the high efficiency compression technology of RGB video sequences are mainly studied.Aiming at the problem of the low resolution,noise inclusion and blurring of the depth edge of the depth camera,a color guided depth restoration algorithm based on depth map smoothness threshold is proposed.And,for the compression performance problem of RGB video sequences,an improved b ilateral adaptive quantization method is proposed in the HEVC quantization coding stage.In order to enhance the quality of deep data,the paper studies the restoration technology of the depth map.We analyze the feasibility of adding the guidance of the color image in the depth restoration process,and propose a strategy for adjusting the weight of depth map items based on the autoregressive model algorithm.The adjustment rule is to perform a different repair on the flat area and the depth edge of the depth map.Then,this paper proposes to use the relative smoothness of the depth map to represent the change of the depth map content.Based on the above work,the paper proposes a repair method of the depth map weight adjustment based on the smoothness threshold.This method uses the calculated the relative smoothness of the depth map as a threshold to modify the guide weight.Qualitative and quantitative experimental results are used to evaluate the proposed algorithm on the RGBD test datasets,and better repair results are obtained.The compression of RGB video in depth camera usually uses the HEVC standard which is based on a hybrid coding framework.In the coding framework,the quantization module is the primary reason for introducing coding distortion.This article focuses on improving the quantization process to raise the performance gain of HEVC video coding.After studying the unified scalar quantization method in the standard and analyzing the improvement direction of quantification,the paper proposes a bilateral adaptive quantification method.This method belongs to the coefficient-level quantification method and includes a three-part algorithm strategy.First,a new quantized scanning strategy is proposed before the quantization coding.Under this scanning method,the arrangement of the transform coefficients is reorganized,and the new coefficient sequence can facilitate the subsequent transform coefficient encoding process.Second,refer to the coefficient value of each transform coefficient itself in the TU,and assign different weight to the transform coefficient to reflect its different contribution in video reconstruction.Third,combine the above strategies to construct a bilateral factor to indirectly modify the quantization step.The proposed algorithm is tested on the official reference encoder HM16.0.Compared with the equal block quantization in the HEVC standard,the coefficient-level quantization algorithm in the paper effectively improves the encoding performance.The algorithm proposed in the paper has been tested and verified in related datasets.The experimental results show that the improved algorithm can meet the requirements of preprocessing and compression coding of the depth camera.